{
 "credit": "Findra (https://findra.dev/)",
 "kind": "Reported by others, not tested by us",
 "license": "CC BY 4.0",
 "license_url": "https://creativecommons.org/licenses/by/4.0/",
 "note": "The numbers were published by the sources named in each row; Findra compiled them. Credit the original source with the figure.",
 "rows": [
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1",
   "gpu_cores": 7,
   "model": "Llama 2 7B",
   "prompt_tps": 108.21,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 7.92
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1",
   "gpu_cores": 7,
   "model": "Llama 2 7B",
   "prompt_tps": 107.81,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 14.19
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1",
   "gpu_cores": 8,
   "model": "Llama 2 7B",
   "prompt_tps": 117.25,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 7.91
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1",
   "gpu_cores": 8,
   "model": "Llama 2 7B",
   "prompt_tps": 117.96,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 14.15
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Pro",
   "gpu_cores": 14,
   "model": "Llama 2 7B",
   "prompt_tps": 262.65,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 12.75
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Pro",
   "gpu_cores": 14,
   "model": "Llama 2 7B",
   "prompt_tps": 235.16,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 21.95
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Pro",
   "gpu_cores": 14,
   "model": "Llama 2 7B",
   "prompt_tps": 232.55,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 35.52
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Pro",
   "gpu_cores": 16,
   "model": "Llama 2 7B",
   "prompt_tps": 302.14,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 12.75
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Pro",
   "gpu_cores": 16,
   "model": "Llama 2 7B",
   "prompt_tps": 270.37,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 22.34
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Pro",
   "gpu_cores": 16,
   "model": "Llama 2 7B",
   "prompt_tps": 266.25,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 36.41
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Max",
   "gpu_cores": 24,
   "model": "Llama 2 7B",
   "prompt_tps": 453.03,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 22.55
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Max",
   "gpu_cores": 24,
   "model": "Llama 2 7B",
   "prompt_tps": 405.87,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 37.81
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Max",
   "gpu_cores": 24,
   "model": "Llama 2 7B",
   "prompt_tps": 400.26,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 54.61
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Max",
   "gpu_cores": 32,
   "model": "Llama 2 7B",
   "prompt_tps": 599.53,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 23.03
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Max",
   "gpu_cores": 32,
   "model": "Llama 2 7B",
   "prompt_tps": 537.37,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 40.2
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Max",
   "gpu_cores": 32,
   "model": "Llama 2 7B",
   "prompt_tps": 530.06,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 61.19
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Ultra",
   "gpu_cores": 48,
   "model": "Llama 2 7B",
   "prompt_tps": 875.81,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 33.92
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Ultra",
   "gpu_cores": 48,
   "model": "Llama 2 7B",
   "prompt_tps": 783.45,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 55.69
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Ultra",
   "gpu_cores": 48,
   "model": "Llama 2 7B",
   "prompt_tps": 772.24,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 74.93
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Ultra",
   "gpu_cores": 64,
   "model": "Llama 2 7B",
   "prompt_tps": 1168.89,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 37.01
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Ultra",
   "gpu_cores": 64,
   "model": "Llama 2 7B",
   "prompt_tps": 1042.95,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 59.87
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M1 Ultra",
   "gpu_cores": 64,
   "model": "Llama 2 7B",
   "prompt_tps": 1030.04,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 83.73
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2",
   "gpu_cores": 8,
   "model": "Llama 2 7B",
   "prompt_tps": 147.27,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 12.18
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2",
   "gpu_cores": 8,
   "model": "Llama 2 7B",
   "prompt_tps": 145.91,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 21.7
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2",
   "gpu_cores": 10,
   "model": "Llama 2 7B",
   "prompt_tps": 201.34,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 6.72
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2",
   "gpu_cores": 10,
   "model": "Llama 2 7B",
   "prompt_tps": 181.4,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 12.21
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2",
   "gpu_cores": 10,
   "model": "Llama 2 7B",
   "prompt_tps": 179.57,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 21.91
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Pro",
   "gpu_cores": 16,
   "model": "Llama 2 7B",
   "prompt_tps": 312.65,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 12.47
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Pro",
   "gpu_cores": 16,
   "model": "Llama 2 7B",
   "prompt_tps": 288.46,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 22.7
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Pro",
   "gpu_cores": 16,
   "model": "Llama 2 7B",
   "prompt_tps": 294.24,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 37.87
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Pro",
   "gpu_cores": 19,
   "model": "Llama 2 7B",
   "prompt_tps": 384.38,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 13.06
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Pro",
   "gpu_cores": 19,
   "model": "Llama 2 7B",
   "prompt_tps": 344.5,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 23.01
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Pro",
   "gpu_cores": 19,
   "model": "Llama 2 7B",
   "prompt_tps": 341.19,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 38.86
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Max",
   "gpu_cores": 30,
   "model": "Llama 2 7B",
   "prompt_tps": 600.46,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 24.16
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Max",
   "gpu_cores": 30,
   "model": "Llama 2 7B",
   "prompt_tps": 540.15,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 39.97
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Max",
   "gpu_cores": 30,
   "model": "Llama 2 7B",
   "prompt_tps": 537.6,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 60.99
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Max",
   "gpu_cores": 38,
   "model": "Llama 2 7B",
   "prompt_tps": 755.67,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 24.65
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Max",
   "gpu_cores": 38,
   "model": "Llama 2 7B",
   "prompt_tps": 677.91,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 41.83
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Max",
   "gpu_cores": 38,
   "model": "Llama 2 7B",
   "prompt_tps": 671.31,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 65.95
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Ultra",
   "gpu_cores": 60,
   "model": "Llama 2 7B",
   "prompt_tps": 1128.59,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 39.86
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Ultra",
   "gpu_cores": 60,
   "model": "Llama 2 7B",
   "prompt_tps": 1003.16,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 62.14
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Ultra",
   "gpu_cores": 60,
   "model": "Llama 2 7B",
   "prompt_tps": 1013.81,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 88.64
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Ultra",
   "gpu_cores": 76,
   "model": "Llama 2 7B",
   "prompt_tps": 1401.85,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 41.02
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Ultra",
   "gpu_cores": 76,
   "model": "Llama 2 7B",
   "prompt_tps": 1248.59,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 66.64
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M2 Ultra",
   "gpu_cores": 76,
   "model": "Llama 2 7B",
   "prompt_tps": 1238.48,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 94.27
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Pro",
   "gpu_cores": 18,
   "model": "Llama 2 7B",
   "prompt_tps": 357.45,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 9.89
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Pro",
   "gpu_cores": 18,
   "model": "Llama 2 7B",
   "prompt_tps": 344.66,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 17.53
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Pro",
   "gpu_cores": 18,
   "model": "Llama 2 7B",
   "prompt_tps": 341.67,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 30.74
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Max",
   "gpu_cores": 30,
   "model": "Llama 2 7B",
   "prompt_tps": 589.41,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 19.54
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Max",
   "gpu_cores": 30,
   "model": "Llama 2 7B",
   "prompt_tps": 566.4,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 34.3
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Max",
   "gpu_cores": 30,
   "model": "Llama 2 7B",
   "prompt_tps": 567.59,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 56.58
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Max",
   "gpu_cores": 40,
   "model": "Llama 2 7B",
   "prompt_tps": 779.17,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 25.09
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Max",
   "gpu_cores": 40,
   "model": "Llama 2 7B",
   "prompt_tps": 757.64,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 42.75
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Max",
   "gpu_cores": 40,
   "model": "Llama 2 7B",
   "prompt_tps": 759.7,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 66.31
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Ultra",
   "gpu_cores": 60,
   "model": "Llama 2 7B",
   "prompt_tps": 1121.8,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 42.24
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Ultra",
   "gpu_cores": 60,
   "model": "Llama 2 7B",
   "prompt_tps": 1085.76,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 63.55
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Ultra",
   "gpu_cores": 60,
   "model": "Llama 2 7B",
   "prompt_tps": 1073.09,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 88.4
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Ultra",
   "gpu_cores": 80,
   "model": "Llama 2 7B",
   "prompt_tps": 1538.34,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 39.78
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Ultra",
   "gpu_cores": 80,
   "model": "Llama 2 7B",
   "prompt_tps": 1487.51,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 63.93
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M3 Ultra",
   "gpu_cores": 80,
   "model": "Llama 2 7B",
   "prompt_tps": 1471.24,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 92.14
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4",
   "gpu_cores": 10,
   "model": "Llama 2 7B",
   "prompt_tps": 230.18,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 7.43
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4",
   "gpu_cores": 10,
   "model": "Llama 2 7B",
   "prompt_tps": 223.64,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 13.54
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4",
   "gpu_cores": 10,
   "model": "Llama 2 7B",
   "prompt_tps": 221.29,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 24.11
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4 Pro",
   "gpu_cores": 16,
   "model": "Llama 2 7B",
   "prompt_tps": 381.14,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 17.19
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4 Pro",
   "gpu_cores": 16,
   "model": "Llama 2 7B",
   "prompt_tps": 367.13,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 30.54
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4 Pro",
   "gpu_cores": 16,
   "model": "Llama 2 7B",
   "prompt_tps": 364.06,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 49.64
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4 Pro",
   "gpu_cores": 20,
   "model": "Llama 2 7B",
   "prompt_tps": 464.48,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 17.18
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4 Pro",
   "gpu_cores": 20,
   "model": "Llama 2 7B",
   "prompt_tps": 449.62,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 30.69
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4 Pro",
   "gpu_cores": 20,
   "model": "Llama 2 7B",
   "prompt_tps": 439.78,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 50.74
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4 Max",
   "gpu_cores": 32,
   "model": "Llama 2 7B",
   "prompt_tps": 736.25,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 24.29
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4 Max",
   "gpu_cores": 32,
   "model": "Llama 2 7B",
   "prompt_tps": 718.56,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 43.87
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4 Max",
   "gpu_cores": 32,
   "model": "Llama 2 7B",
   "prompt_tps": 713.93,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 69.95
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4 Max",
   "gpu_cores": 40,
   "model": "Llama 2 7B",
   "prompt_tps": 922.83,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 31.64
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4 Max",
   "gpu_cores": 40,
   "model": "Llama 2 7B",
   "prompt_tps": 891.94,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 54.05
  },
  {
   "backend": "Metal",
   "build": "8e672ef",
   "chip": "Apple M4 Max",
   "gpu_cores": 40,
   "model": "Llama 2 7B",
   "prompt_tps": 885.68,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 83.06
  },
  {
   "backend": "Metal",
   "build": "60081bb",
   "chip": "Apple M5 Pro",
   "gpu_cores": 16,
   "model": "Llama 2 7B",
   "prompt_tps": 1310.78,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 21.39
  },
  {
   "backend": "Metal",
   "build": "60081bb",
   "chip": "Apple M5 Pro",
   "gpu_cores": 16,
   "model": "Llama 2 7B",
   "prompt_tps": 1302.59,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 38.94
  },
  {
   "backend": "Metal",
   "build": "60081bb",
   "chip": "Apple M5 Pro",
   "gpu_cores": 16,
   "model": "Llama 2 7B",
   "prompt_tps": 1340.44,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 67.59
  },
  {
   "backend": "Metal",
   "build": "c1d0e7a",
   "chip": "Apple M5 Pro",
   "gpu_cores": 20,
   "model": "Llama 2 7B",
   "prompt_tps": 1588.78,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 21.55
  },
  {
   "backend": "Metal",
   "build": "c1d0e7a",
   "chip": "Apple M5 Pro",
   "gpu_cores": 20,
   "model": "Llama 2 7B",
   "prompt_tps": 1553.86,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 38.92
  },
  {
   "backend": "Metal",
   "build": "c1d0e7a",
   "chip": "Apple M5 Pro",
   "gpu_cores": 20,
   "model": "Llama 2 7B",
   "prompt_tps": 1620.64,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 66.33
  },
  {
   "backend": "Metal",
   "build": "c1d0e7a",
   "chip": "Apple M5 Max",
   "gpu_cores": 40,
   "model": "Llama 2 7B",
   "prompt_tps": 3158.49,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 37.11
  },
  {
   "backend": "Metal",
   "build": "c1d0e7a",
   "chip": "Apple M5 Max",
   "gpu_cores": 40,
   "model": "Llama 2 7B",
   "prompt_tps": 3143.81,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 72.42
  },
  {
   "backend": "Metal",
   "build": "c1d0e7a",
   "chip": "Apple M5 Max",
   "gpu_cores": 40,
   "model": "Llama 2 7B",
   "prompt_tps": 3219.99,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 119.92
  },
  {
   "backend": "Metal",
   "build": "c1d0e7a",
   "chip": "Apple M5 Ultra",
   "gpu_cores": 80,
   "model": "Llama 2 7B",
   "prompt_tps": 5207.14,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 73.6
  },
  {
   "backend": "Metal",
   "build": "c1d0e7a",
   "chip": "Apple M5 Ultra",
   "gpu_cores": 80,
   "model": "Llama 2 7B",
   "prompt_tps": 4910.55,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 119.01
  },
  {
   "backend": "Metal",
   "build": "c1d0e7a",
   "chip": "Apple M5 Ultra",
   "gpu_cores": 80,
   "model": "Llama 2 7B",
   "prompt_tps": 4944.66,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 179.1
  },
  {
   "backend": "Metal",
   "build": "c1d0e7a",
   "chip": "Apple M6",
   "gpu_cores": 8,
   "model": "Llama 2 7B",
   "prompt_tps": 952.53,
   "quant": "F16",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 11.59
  },
  {
   "backend": "Metal",
   "build": "c1d0e7a",
   "chip": "Apple M6",
   "gpu_cores": 8,
   "model": "Llama 2 7B",
   "prompt_tps": 970.46,
   "quant": "Q8_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 21.01
  },
  {
   "backend": "Metal",
   "build": "c1d0e7a",
   "chip": "Apple M6",
   "gpu_cores": 8,
   "model": "Llama 2 7B",
   "prompt_tps": 988.93,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 4167",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "write_tps": 35.74
  },
  {
   "backend": "CUDA",
   "build": "8cf6b42",
   "chip": "RTX 5090",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 14073.41,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 290.02
  },
  {
   "backend": "CUDA",
   "build": "79c1160",
   "chip": "RTX PRO 6000 Blackwell",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 14854.63,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 274.2
  },
  {
   "backend": "CUDA",
   "build": "5143fa8",
   "chip": "H100 80 GB",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 9918.34,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 267.81
  },
  {
   "backend": "CUDA",
   "build": "5143fa8",
   "chip": "A100 80 GB",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4849.53,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 190.88
  },
  {
   "backend": "CUDA",
   "build": "79c1160",
   "chip": "RTX 4090 D",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 10293.86,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 189.33
  },
  {
   "backend": "CUDA",
   "build": "2241453",
   "chip": "RTX 4090",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 11992.7,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 186.21
  },
  {
   "backend": "CUDA",
   "build": "8a4280c",
   "chip": "RTX 5080",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 8297.36,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 181.99
  },
  {
   "backend": "CUDA",
   "build": "933414c",
   "chip": "RTX 5070 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 6952.38,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 176.85
  },
  {
   "backend": "CUDA",
   "build": "b8e09f0",
   "chip": "RTX 6000 Ada",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 9229.23,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 176.07
  },
  {
   "backend": "CUDA",
   "build": "9c35706",
   "chip": "RTX 3090 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 6567.49,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 171.19
  },
  {
   "backend": "CUDA",
   "build": "c76b420",
   "chip": "RTX 3090",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5174.69,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 158.16
  },
  {
   "backend": "CUDA",
   "build": "29f1482",
   "chip": "RTX 5090 Laptop",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 6667.06,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 156.49
  },
  {
   "backend": "CUDA",
   "build": "ee09828",
   "chip": "L40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 8870.49,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 152.01
  },
  {
   "backend": "CUDA",
   "build": "81086cd",
   "chip": "RTX 4080 SUPER",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 8125.15,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 148.33
  },
  {
   "backend": "CUDA",
   "build": "20638e4",
   "chip": "RTX 4080",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 8031.64,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 142.49
  },
  {
   "backend": "CUDA",
   "build": "9c35706",
   "chip": "RTX 3080",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5013.86,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 139.65
  },
  {
   "backend": "CUDA",
   "build": "4795c91",
   "chip": "RTX A6000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4913.93,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 138.73
  },
  {
   "backend": "CUDA",
   "build": "9c35706",
   "chip": "RTX 4070 Ti SUPER",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 6924.53,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 132.26
  },
  {
   "backend": "CUDA",
   "build": "7d77f07",
   "chip": "RTX PRO 4000 Blackwell",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4992.83,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 131.66
  },
  {
   "backend": "CUDA",
   "build": "e5155e6",
   "chip": "RTX A5000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4028.16,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 130.07
  },
  {
   "backend": "CUDA",
   "build": "51f5a45",
   "chip": "Tesla V100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3042.64,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 129.08
  },
  {
   "backend": "CUDA",
   "build": "",
   "chip": "RTX 5070",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5184.75,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 127.54
  },
  {
   "backend": "CUDA",
   "build": "3470a5c",
   "chip": "A40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4609.01,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 124.11
  },
  {
   "backend": "CUDA",
   "build": "583cb83",
   "chip": "A30",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2767.1,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 124.81
  },
  {
   "backend": "CUDA",
   "build": "e56abd2",
   "chip": "Titan V",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2617.46,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 108.79
  },
  {
   "backend": "CUDA",
   "build": "9c35706",
   "chip": "RTX 2080 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2890.66,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 107.51
  },
  {
   "backend": "CUDA",
   "build": "b8e09f0",
   "chip": "Quadro RTX 6000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2751.18,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 102.77
  },
  {
   "backend": "CUDA",
   "build": "b8e09f0",
   "chip": "Quadro RTX 8000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2709.95,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 102.68
  },
  {
   "backend": "CUDA",
   "build": "5cdb27e",
   "chip": "RTX A4500",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2827.2,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 97.32
  },
  {
   "backend": "CUDA",
   "build": "89d10295",
   "chip": "RTX 5060 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3737.25,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 90.94
  },
  {
   "backend": "CUDA",
   "build": "bc07349",
   "chip": "RTX 2070 SUPER",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2088.34,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 88.06
  },
  {
   "backend": "CUDA",
   "build": "65349f2",
   "chip": "RTX A4000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2684.06,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 83.77
  },
  {
   "backend": "CUDA",
   "build": "c4510dc",
   "chip": "Titan Xp",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1154.96,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 76.08
  },
  {
   "backend": "CUDA",
   "build": "baa9255",
   "chip": "RTX 3060",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2137.5,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 75.57
  },
  {
   "backend": "CUDA",
   "build": "7d77f07",
   "chip": "Quadro RTX 4000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1536.89,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 65.62
  },
  {
   "backend": "CUDA",
   "build": "89d10295",
   "chip": "RTX 4060 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3394.63,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 63.86
  },
  {
   "backend": "CUDA",
   "build": "9c35706",
   "chip": "GTX 1080 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1084.41,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 62.49
  },
  {
   "backend": "CUDA",
   "build": "a74a0d6",
   "chip": "RTX A4000 Ada",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2779.77,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 61.83
  },
  {
   "backend": "CUDA",
   "build": "5c0eb5e",
   "chip": "RTX 2060 SUPER",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1420.24,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 60.04
  },
  {
   "backend": "CUDA",
   "build": "b8372ee",
   "chip": "Tesla P100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 760.8,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 58.35
  },
  {
   "backend": "CUDA",
   "build": "5acd455",
   "chip": "DGX Spark",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3062.31,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 57.21
  },
  {
   "backend": "CUDA",
   "build": "c76b420",
   "chip": "Tesla P40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1007.42,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 54.74
  },
  {
   "backend": "CUDA",
   "build": "756cfea",
   "chip": "RTX 2000 Ada",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1956.22,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 50.62
  },
  {
   "backend": "CUDA",
   "build": "d32e03f",
   "chip": "Tesla T4",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1219.06,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 46.38
  },
  {
   "backend": "CUDA",
   "build": "d79d8f3",
   "chip": "RTX 4050 Laptop",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1725.85,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 43.72
  },
  {
   "backend": "CUDA",
   "build": "9515c61",
   "chip": "GTX 1660",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 148.91,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 41.35
  },
  {
   "backend": "CUDA",
   "build": "97d5117",
   "chip": "Tesla M40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 282.65,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 38.04
  },
  {
   "backend": "CUDA",
   "build": "79c1160",
   "chip": "GTX 1070 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 714.44,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 37.82
  },
  {
   "backend": "CUDA",
   "build": "c1b1876",
   "chip": "Jetson AGX Orin",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 991.31,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 33.58
  },
  {
   "backend": "CUDA",
   "build": "c76b420",
   "chip": "Tesla P4",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 514.53,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 33.29
  },
  {
   "backend": "CUDA",
   "build": "5fd160b",
   "chip": "P106-100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 406.94,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 30.4
  },
  {
   "backend": "CUDA",
   "build": "5fd160b",
   "chip": "GTX 1060",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 416.85,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 27.79
  },
  {
   "backend": "CUDA",
   "build": "f6da8cb",
   "chip": "Quadro T1000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 79.44,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 27.82
  },
  {
   "backend": "CUDA",
   "build": "baa9255",
   "chip": "Quadro P2000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 309.3,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 23.63
  },
  {
   "backend": "CUDA",
   "build": "1e74897",
   "chip": "Quadro P1000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 183.4,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 13.99
  },
  {
   "backend": "CUDA",
   "build": "32732f2",
   "chip": "Tesla K80",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 133.14,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 13.8
  },
  {
   "backend": "CUDA",
   "build": "8cf6b42",
   "chip": "RTX 5090",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 14970.15,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 300.4
  },
  {
   "backend": "CUDA",
   "build": "5143fa8",
   "chip": "RTX PRO 6000 Blackwell",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 16618.98,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 281.11
  },
  {
   "backend": "CUDA",
   "build": "5143fa8",
   "chip": "H100 80 GB",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 11263.29,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 280.74
  },
  {
   "backend": "CUDA",
   "build": "5143fa8",
   "chip": "A100 80 GB",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5285.96,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 200.9
  },
  {
   "backend": "CUDA",
   "build": "79c1160",
   "chip": "RTX 4090 D",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 12506.97,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 191.57
  },
  {
   "backend": "CUDA",
   "build": "2241453",
   "chip": "RTX 4090",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 14770.63,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 188.96
  },
  {
   "backend": "CUDA",
   "build": "8a4280c",
   "chip": "RTX 5080",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 9487.7,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 184.68
  },
  {
   "backend": "CUDA",
   "build": "933414c",
   "chip": "RTX 5070 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 8419.56,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 182.43
  },
  {
   "backend": "CUDA",
   "build": "b8e09f0",
   "chip": "RTX 6000 Ada",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 10576.85,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 179.47
  },
  {
   "backend": "CUDA",
   "build": "9c35706",
   "chip": "RTX 3090 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 6924.01,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 172.26
  },
  {
   "backend": "CUDA",
   "build": "becc481",
   "chip": "RTX PRO 4500 Blackwell",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 7251.66,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 168.9
  },
  {
   "backend": "CUDA",
   "build": "c76b420",
   "chip": "RTX 3090",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5560.06,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 161.89
  },
  {
   "backend": "CUDA",
   "build": "29f1482",
   "chip": "RTX 5090 Laptop",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 7641.89,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 158.14
  },
  {
   "backend": "CUDA",
   "build": "ee09828",
   "chip": "L40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 10097.64,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 153.76
  },
  {
   "backend": "CUDA",
   "build": "81086cd",
   "chip": "RTX 4080 SUPER",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 9439.01,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 147.48
  },
  {
   "backend": "CUDA",
   "build": "20638e4",
   "chip": "RTX 4080",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 9205.93,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 143.47
  },
  {
   "backend": "CUDA",
   "build": "4795c91",
   "chip": "RTX A6000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5662.39,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 144.87
  },
  {
   "backend": "CUDA",
   "build": "9c35706",
   "chip": "RTX 3080",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5569.56,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 139.95
  },
  {
   "backend": "CUDA",
   "build": "7d77f07",
   "chip": "RTX PRO 4000 Blackwell",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5674.44,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 136.38
  },
  {
   "backend": "CUDA",
   "build": "e5155e6",
   "chip": "RTX A5000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4552.15,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 135.83
  },
  {
   "backend": "CUDA",
   "build": "51f5a45",
   "chip": "Tesla V100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2973.78,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 134.76
  },
  {
   "backend": "CUDA",
   "build": "9c35706",
   "chip": "RTX 4070 Ti SUPER",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 7612.32,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 132.85
  },
  {
   "backend": "CUDA",
   "build": "",
   "chip": "RTX 5070",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5783.44,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 128.21
  },
  {
   "backend": "CUDA",
   "build": "3470a5c",
   "chip": "A40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5256.38,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 126.24
  },
  {
   "backend": "CUDA",
   "build": "583cb83",
   "chip": "A30",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3068.72,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 131.93
  },
  {
   "backend": "CUDA",
   "build": "e56abd2",
   "chip": "Titan V",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2481.25,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 112.17
  },
  {
   "backend": "CUDA",
   "build": "9c35706",
   "chip": "RTX 2080 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3107.61,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 109.17
  },
  {
   "backend": "CUDA",
   "build": "b8e09f0",
   "chip": "Quadro RTX 6000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3053.96,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 104.38
  },
  {
   "backend": "CUDA",
   "build": "b8e09f0",
   "chip": "Quadro RTX 8000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3052.35,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 103.63
  },
  {
   "backend": "CUDA",
   "build": "5cdb27e",
   "chip": "RTX A4500",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3453.1,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 103.0
  },
  {
   "backend": "CUDA",
   "build": "89d10295",
   "chip": "RTX 5060 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4195.53,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 93.46
  },
  {
   "backend": "CUDA",
   "build": "bc07349",
   "chip": "RTX 2070 SUPER",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2293.29,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 87.71
  },
  {
   "backend": "CUDA",
   "build": "65349f2",
   "chip": "RTX A4000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2807.83,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 85.17
  },
  {
   "backend": "CUDA",
   "build": "baa9255",
   "chip": "RTX 3060",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2407.67,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 76.92
  },
  {
   "backend": "CUDA",
   "build": "c4510dc",
   "chip": "Titan Xp",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1218.12,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 73.84
  },
  {
   "backend": "CUDA",
   "build": "7d77f07",
   "chip": "Quadro RTX 4000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1662.8,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 67.62
  },
  {
   "backend": "CUDA",
   "build": "89d10295",
   "chip": "RTX 4060 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3803.45,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 64.03
  },
  {
   "backend": "CUDA",
   "build": "a74a0d6",
   "chip": "RTX A4000 Ada",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3171.86,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 61.37
  },
  {
   "backend": "CUDA",
   "build": "b8372ee",
   "chip": "Tesla P100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 787.36,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 61.99
  },
  {
   "backend": "CUDA",
   "build": "9c35706",
   "chip": "GTX 1080 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1138.14,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 61.38
  },
  {
   "backend": "CUDA",
   "build": "5c0eb5e",
   "chip": "RTX 2060 SUPER",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1563.77,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 61.13
  },
  {
   "backend": "CUDA",
   "build": "5acd455",
   "chip": "DGX Spark",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3661.37,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 56.74
  },
  {
   "backend": "CUDA",
   "build": "c76b420",
   "chip": "Tesla P40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1079.66,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 53.73
  },
  {
   "backend": "CUDA",
   "build": "756cfea",
   "chip": "RTX 2000 Ada",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2250.14,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 50.71
  },
  {
   "backend": "CUDA",
   "build": "d32e03f",
   "chip": "Tesla T4",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1309.73,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 44.03
  },
  {
   "backend": "CUDA",
   "build": "9515c61",
   "chip": "GTX 1660",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 154.45,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 41.43
  },
  {
   "backend": "CUDA",
   "build": "97d5117",
   "chip": "Tesla M40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 290.17,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 39.98
  },
  {
   "backend": "CUDA",
   "build": "79c1160",
   "chip": "GTX 1070 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 790.52,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 37.87
  },
  {
   "backend": "CUDA",
   "build": "c1b1876",
   "chip": "Jetson AGX Orin",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1171.96,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 35.88
  },
  {
   "backend": "CUDA",
   "build": "c76b420",
   "chip": "Tesla P4",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 529.53,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 33.12
  },
  {
   "backend": "CUDA",
   "build": "5fd160b",
   "chip": "P106-100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 438.49,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 30.64
  },
  {
   "backend": "CUDA",
   "build": "5fd160b",
   "chip": "GTX 1060",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 446.19,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 28.18
  },
  {
   "backend": "CUDA",
   "build": "baa9255",
   "chip": "Quadro P2000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 311.55,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 23.76
  },
  {
   "backend": "CUDA",
   "build": "32732f2",
   "chip": "Tesla K80",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 133.36,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 14.27
  },
  {
   "backend": "CUDA",
   "build": "1e74897",
   "chip": "Quadro P1000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 173.82,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15013",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "write_tps": 13.65
  },
  {
   "backend": "ROCm (HIP)",
   "build": "ee3a9fc",
   "chip": "Instinct MI300X",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 11476.4,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 232.92
  },
  {
   "backend": "ROCm (HIP)",
   "build": "2f0c2db",
   "chip": "RX 7900 XTX",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3552.27,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 167.11
  },
  {
   "backend": "ROCm (HIP)",
   "build": "8160b38",
   "chip": "Instinct MI210",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2486.22,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 124.51
  },
  {
   "backend": "ROCm (HIP)",
   "build": "8160b38",
   "chip": "Pro W7900",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3213.17,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 121.18
  },
  {
   "backend": "ROCm (HIP)",
   "build": "1e15bfd",
   "chip": "RX 7900 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3098.38,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 116.15
  },
  {
   "backend": "ROCm (HIP)",
   "build": "d0660f2",
   "chip": "RX 9070",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2381.77,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 114.48
  },
  {
   "backend": "ROCm (HIP)",
   "build": "9c35706",
   "chip": "Instinct MI100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2732.83,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 110.48
  },
  {
   "backend": "ROCm (HIP)",
   "build": "583cb83",
   "chip": "RX 9070 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5055.19,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 101.27
  },
  {
   "backend": "ROCm (HIP)",
   "build": "00131d6",
   "chip": "RX 7800 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2151.81,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 100.94
  },
  {
   "backend": "ROCm (HIP)",
   "build": "97d5117",
   "chip": "Instinct MI50",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1057.24,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 98.95
  },
  {
   "backend": "ROCm (HIP)",
   "build": "6fa3b55",
   "chip": "RX 7900 GRE",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1456.98,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 96.07
  },
  {
   "backend": "ROCm (HIP)",
   "build": "bd4ef13",
   "chip": "AI PRO R9700",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4443.54,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 93.84
  },
  {
   "backend": "ROCm (HIP)",
   "build": "504af20",
   "chip": "Instinct MI60",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1289.11,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 91.46
  },
  {
   "backend": "ROCm (HIP)",
   "build": "a972fae",
   "chip": "RX 6900 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1889.84,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 88.49
  },
  {
   "backend": "ROCm (HIP)",
   "build": "2739a71",
   "chip": "Pro VII",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1064.99,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 87.45
  },
  {
   "backend": "ROCm (HIP)",
   "build": "79c1160",
   "chip": "RX 6800 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1447.07,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 83.92
  },
  {
   "backend": "ROCm (HIP)",
   "build": "5c0eb5e",
   "chip": "Pro V620",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1803.65,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 74.66
  },
  {
   "backend": "ROCm (HIP)",
   "build": "a0e13dc",
   "chip": "RX 9060 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1419.67,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 67.58
  },
  {
   "backend": "ROCm (HIP)",
   "build": "c05e8c9",
   "chip": "RX 5700 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 354.17,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 67.55
  },
  {
   "backend": "ROCm (HIP)",
   "build": "2739a71",
   "chip": "Instinct MI25",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 409.83,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 63.94
  },
  {
   "backend": "ROCm (HIP)",
   "build": "e60f241",
   "chip": "AI Max+ 395",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 911.36,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 50.01
  },
  {
   "backend": "ROCm (HIP)",
   "build": "9c35706",
   "chip": "RX 7600 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1099.64,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 48.58
  },
  {
   "backend": "ROCm (HIP)",
   "build": "ec428b0",
   "chip": "RX Vega 64",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 240.68,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 48.46
  },
  {
   "backend": "ROCm (HIP)",
   "build": "1d0125b",
   "chip": "Radeon 8060S",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 351.36,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 47.97
  },
  {
   "backend": "ROCm (HIP)",
   "build": "c55d53a",
   "chip": "Radeon 880M",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 163.25,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 12.97
  },
  {
   "backend": "ROCm (HIP)",
   "build": "ee3a9fc",
   "chip": "Instinct MI300X",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 11945.97,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 218.53
  },
  {
   "backend": "ROCm (HIP)",
   "build": "2f0c2db",
   "chip": "RX 7900 XTX",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3874.25,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 170.12
  },
  {
   "backend": "ROCm (HIP)",
   "build": "8160b38",
   "chip": "Instinct MI210",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2571.82,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 130.18
  },
  {
   "backend": "ROCm (HIP)",
   "build": "8160b38",
   "chip": "Pro W7900",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3472.86,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 127.43
  },
  {
   "backend": "ROCm (HIP)",
   "build": "d0660f2",
   "chip": "RX 9070",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2452.68,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 115.32
  },
  {
   "backend": "ROCm (HIP)",
   "build": "1e15bfd",
   "chip": "RX 7900 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3261.75,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 112.3
  },
  {
   "backend": "ROCm (HIP)",
   "build": "9c35706",
   "chip": "Instinct MI100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2755.0,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 104.71
  },
  {
   "backend": "ROCm (HIP)",
   "build": "97d5117",
   "chip": "Instinct MI50",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1129.43,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 105.82
  },
  {
   "backend": "ROCm (HIP)",
   "build": "bd4ef13",
   "chip": "AI PRO R9700",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4773.07,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 97.98
  },
  {
   "backend": "ROCm (HIP)",
   "build": "6fa3b55",
   "chip": "RX 7900 GRE",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1598.79,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 97.53
  },
  {
   "backend": "ROCm (HIP)",
   "build": "583cb83",
   "chip": "RX 9070 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4903.51,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 97.28
  },
  {
   "backend": "ROCm (HIP)",
   "build": "00131d6",
   "chip": "RX 7800 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2304.63,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 95.99
  },
  {
   "backend": "ROCm (HIP)",
   "build": "a972fae",
   "chip": "RX 6900 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1948.31,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 85.04
  },
  {
   "backend": "ROCm (HIP)",
   "build": "5c0eb5e",
   "chip": "Pro V620",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1256.86,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 70.83
  },
  {
   "backend": "ROCm (HIP)",
   "build": "a0e13dc",
   "chip": "RX 9060 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1479.27,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 65.42
  },
  {
   "backend": "ROCm (HIP)",
   "build": "c05e8c9",
   "chip": "RX 5700 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 314.17,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 62.02
  },
  {
   "backend": "ROCm (HIP)",
   "build": "e60f241",
   "chip": "AI Max+ 395",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1003.53,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 49.87
  },
  {
   "backend": "ROCm (HIP)",
   "build": "1d0125b",
   "chip": "Radeon 8060S",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 366.08,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 48.97
  },
  {
   "backend": "ROCm (HIP)",
   "build": "9c35706",
   "chip": "RX 7600 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1199.16,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 47.65
  },
  {
   "backend": "ROCm (HIP)",
   "build": "ec428b0",
   "chip": "RX Vega 64",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 153.17,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 42.46
  },
  {
   "backend": "ROCm (HIP)",
   "build": "c55d53a",
   "chip": "Radeon 880M",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 213.31,
   "quant": "Q4_0",
   "retrieved": "2026-10-10",
   "source": "llama.cpp discussion 15021",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "write_tps": 16.16
  },
  {
   "backend": "Vulkan",
   "build": "ca71fb9",
   "chip": "RTX 5090",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 10381.64,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 263.63
  },
  {
   "backend": "Vulkan",
   "build": "4ae88d0",
   "chip": "RTX 4090",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 9452.03,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 187.97
  },
  {
   "backend": "Vulkan",
   "build": "f6b533d89",
   "chip": "RTX 5080",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 7444.99,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 185.1
  },
  {
   "backend": "Vulkan",
   "build": "304665f",
   "chip": "RX 7900 XTX",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3726.99,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 182.63
  },
  {
   "backend": "Vulkan",
   "build": "d05fe1d",
   "chip": "RTX 3090",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4666.15,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 164.05
  },
  {
   "backend": "Vulkan",
   "build": "2257758",
   "chip": "A100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 6389.86,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 160.78
  },
  {
   "backend": "Vulkan",
   "build": "4bc272f",
   "chip": "RTX 5090 Mobile",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5546.39,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 149.41
  },
  {
   "backend": "Vulkan",
   "build": "81086cd",
   "chip": "RTX 4080 Super",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 7101.18,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 147.13
  },
  {
   "backend": "Vulkan",
   "build": "dd1ea5243",
   "chip": "AI Pro R9700",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5609.82,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 145.67
  },
  {
   "backend": "Vulkan",
   "build": "7c7d6ce",
   "chip": "RTX 3080",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4287.11,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 139.15
  },
  {
   "backend": "Vulkan",
   "build": "4ae88d0",
   "chip": "RTX A5000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3641.55,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 139.89
  },
  {
   "backend": "Vulkan",
   "build": "e9fd8dc",
   "chip": "RX 9070 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5036.04,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 137.11
  },
  {
   "backend": "Vulkan",
   "build": "d13d0f6",
   "chip": "RTX 5070 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 6213.63,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 135.63
  },
  {
   "backend": "Vulkan",
   "build": "7d77f07",
   "chip": "Tesla V100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1391.39,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 129.58
  },
  {
   "backend": "Vulkan",
   "build": "4ae88d0",
   "chip": "RTX 4070 Ti Super",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 6099.18,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 129.45
  },
  {
   "backend": "Vulkan",
   "build": "71e74a3",
   "chip": "RX 7900 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2941.58,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 123.18
  },
  {
   "backend": "Vulkan",
   "build": "21c17b5",
   "chip": "RX 9070",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3164.1,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 119.71
  },
  {
   "backend": "Vulkan",
   "build": "4fdbc1e",
   "chip": "RX 7800 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2017.33,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 118.27
  },
  {
   "backend": "Vulkan",
   "build": "4b2a477",
   "chip": "RX 7900 GRE",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2336.31,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 116.11
  },
  {
   "backend": "Vulkan",
   "build": "2d451c8",
   "chip": "Apple M3 Ultra",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1116.83,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 115.54
  },
  {
   "backend": "Vulkan",
   "build": "b863507",
   "chip": "Intel Arc Pro B70",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3379.0,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 112.02
  },
  {
   "backend": "Vulkan",
   "build": "583cb83",
   "chip": "A30",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3290.7,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 111.48
  },
  {
   "backend": "Vulkan",
   "build": "516a4ca",
   "chip": "RTX 4070 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4981.44,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 110.53
  },
  {
   "backend": "Vulkan",
   "build": "c945aaa",
   "chip": "RTX 4070 Super",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4608.2,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 108.74
  },
  {
   "backend": "Vulkan",
   "build": "3af34b9",
   "chip": "Instinct MI50",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1119.55,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 108.51
  },
  {
   "backend": "Vulkan",
   "build": "a972fae",
   "chip": "RX 6900 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1901.2,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 108.0
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "Radeon Pro VII",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 912.47,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 106.03
  },
  {
   "backend": "Vulkan",
   "build": "e56abd2",
   "chip": "Titan V",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 796.29,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 105.06
  },
  {
   "backend": "Vulkan",
   "build": "f1cee99",
   "chip": "RX 6800",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1593.97,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 101.47
  },
  {
   "backend": "Vulkan",
   "build": "77d6ae4",
   "chip": "Radeon VII",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1059.14,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 101.19
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "RX 6800 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1752.92,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 100.32
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "RTX 2080 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1888.24,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 97.58
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "Radeon Pro W6800X Duo",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 687.71,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 94.82
  },
  {
   "backend": "Vulkan",
   "build": "89f10ba",
   "chip": "RTX 5060 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3460.92,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 93.51
  },
  {
   "backend": "Vulkan",
   "build": "9a483999",
   "chip": "RTX 4070",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3179.37,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 92.29
  },
  {
   "backend": "Vulkan",
   "build": "b42c7fa5b",
   "chip": "RTX 3060 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2361.43,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 86.74
  },
  {
   "backend": "Vulkan",
   "build": "13b4548",
   "chip": "Radeon Pro W6800X",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 510.8,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 86.47
  },
  {
   "backend": "Vulkan",
   "build": "6d75883",
   "chip": "RX 6700 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1051.2,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 83.88
  },
  {
   "backend": "Vulkan",
   "build": "228f34c",
   "chip": "RX 6750 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1040.58,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 81.98
  },
  {
   "backend": "Vulkan",
   "build": "03d4698",
   "chip": "Radeon Pro V620",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1595.32,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 81.78
  },
  {
   "backend": "Vulkan",
   "build": "1b8fb815",
   "chip": "RTX 3070",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2113.02,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 78.71
  },
  {
   "backend": "Vulkan",
   "build": "504af20",
   "chip": "Instinct MI60",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 369.26,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 78.16
  },
  {
   "backend": "Vulkan",
   "build": "92c0b38",
   "chip": "RTX 3060",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1815.7,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 75.94
  },
  {
   "backend": "Vulkan",
   "build": "1ece0cb6",
   "chip": "Apple M4 Max",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 724.77,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 75.02
  },
  {
   "backend": "Vulkan",
   "build": "7f76692",
   "chip": "Tesla T10",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1692.7,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 75.01
  },
  {
   "backend": "Vulkan",
   "build": "f5245b5",
   "chip": "RTX A4000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2248.14,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 73.74
  },
  {
   "backend": "Vulkan",
   "build": "f1cee99",
   "chip": "RX 9060 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2844.8,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 71.07
  },
  {
   "backend": "Vulkan",
   "build": "4fdbc1e",
   "chip": "RX 5700 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 529.69,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 70.73
  },
  {
   "backend": "Vulkan",
   "build": "7f76692",
   "chip": "Intel Arc B580",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 620.94,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 70.14
  },
  {
   "backend": "Vulkan",
   "build": "9da3dcd",
   "chip": "Radeon Pro V540",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 583.88,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 69.64
  },
  {
   "backend": "Vulkan",
   "build": "23bc779",
   "chip": "Radeon Pro W5700",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 449.85,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 68.55
  },
  {
   "backend": "Vulkan",
   "build": "516a4ca",
   "chip": "Intel Arc Pro B60",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 522.36,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 68.55
  },
  {
   "backend": "Vulkan",
   "build": "ac2557c",
   "chip": "GTX 1080 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 585.54,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 67.75
  },
  {
   "backend": "Vulkan",
   "build": "b7552cf",
   "chip": "RTX 2070 Super",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1199.13,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 64.64
  },
  {
   "backend": "Vulkan",
   "build": "ceff6bb",
   "chip": "RTX 3070 Mobile",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1689.4,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 63.64
  },
  {
   "backend": "Vulkan",
   "build": "5dcb711",
   "chip": "RX Vega 64",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 575.06,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 63.43
  },
  {
   "backend": "Vulkan",
   "build": "eec1e33",
   "chip": "Tesla P100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 678.14,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 63.16
  },
  {
   "backend": "Vulkan",
   "build": "ac2557c",
   "chip": "P102-100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 510.62,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 62.83
  },
  {
   "backend": "Vulkan",
   "build": "5886f4f54",
   "chip": "AMD BC-250",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 370.66,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 62.32
  },
  {
   "backend": "Vulkan",
   "build": "dbb852b",
   "chip": "RX 6650 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1029.52,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 62.14
  },
  {
   "backend": "Vulkan",
   "build": "a5c07dc",
   "chip": "RTX 4060 Mobile",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2135.66,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 59.53
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "Tesla P40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 488.06,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 59.36
  },
  {
   "backend": "Vulkan",
   "build": "657e011",
   "chip": "GB10",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2789.51,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 58.51
  },
  {
   "backend": "Vulkan",
   "build": "91f8c9c",
   "chip": "RX 7600",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1167.2,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 58.03
  },
  {
   "backend": "Vulkan",
   "build": "b43556e",
   "chip": "GTX 1660 Ti Mobile",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 511.67,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 56.6
  },
  {
   "backend": "Vulkan",
   "build": "2739a71",
   "chip": "Instinct MI25",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 439.42,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 54.69
  },
  {
   "backend": "Vulkan",
   "build": "091592d",
   "chip": "RX 6600 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 574.65,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 53.92
  },
  {
   "backend": "Vulkan",
   "build": "01d8eaa",
   "chip": "RX 7600 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 840.85,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 53.02
  },
  {
   "backend": "Vulkan",
   "build": "a69d54f",
   "chip": "Intel Arc A770",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1073.85,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 52.56
  },
  {
   "backend": "Vulkan",
   "build": "eec1e33",
   "chip": "AMD FirePro S9300 x2",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 247.26,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 51.86
  },
  {
   "backend": "Vulkan",
   "build": "fb4b273",
   "chip": "AMD Ryzen AI Max+ 300 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1487.08,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 51.81
  },
  {
   "backend": "Vulkan",
   "build": "86a9c79",
   "chip": "RX 6600",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 758.5,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 50.57
  },
  {
   "backend": "Vulkan",
   "build": "92c0b38",
   "chip": "RX Vega 56",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 439.87,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 50.23
  },
  {
   "backend": "Vulkan",
   "build": "7f76692",
   "chip": "Intel Arc B570",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 913.95,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 49.64
  },
  {
   "backend": "Vulkan",
   "build": "dbb3a47",
   "chip": "RTX 3060 Mobile",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1059.76,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 49.03
  },
  {
   "backend": "Vulkan",
   "build": "8e6f8bc8",
   "chip": "RX 6800M",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 861.99,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 48.71
  },
  {
   "backend": "Vulkan",
   "build": "fe5b78c",
   "chip": "RX 6600M",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 605.59,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 48.21
  },
  {
   "backend": "Vulkan",
   "build": "eeee367de",
   "chip": "Intel Arc A770M",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 875.92,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 47.69
  },
  {
   "backend": "Vulkan",
   "build": "aa46bda",
   "chip": "Intel Arc Pro B50",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1186.45,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 46.3
  },
  {
   "backend": "Vulkan",
   "build": "eec1e33",
   "chip": "P104-100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 311.9,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 46.18
  },
  {
   "backend": "Vulkan",
   "build": "b1afcab",
   "chip": "RTX A2000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1245.19,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 45.52
  },
  {
   "backend": "Vulkan",
   "build": "b9ab0a4",
   "chip": "RX 7600M XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 459.39,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 45.28
  },
  {
   "backend": "Vulkan",
   "build": "9da3dcd",
   "chip": "Radeon Pro V340",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 375.41,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 45.16
  },
  {
   "backend": "Vulkan",
   "build": "860a9e4",
   "chip": "GTX 1070 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 297.5,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 42.86
  },
  {
   "backend": "Vulkan",
   "build": "c1b1876",
   "chip": "Intel Arc A750",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1075.94,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 42.66
  },
  {
   "backend": "Vulkan",
   "build": "d79d8f3",
   "chip": "RTX 4050 Mobile",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1154.28,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 41.89
  },
  {
   "backend": "Vulkan",
   "build": "eec1e33",
   "chip": "GTX 1070",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 321.57,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 41.48
  },
  {
   "backend": "Vulkan",
   "build": "32e41fa5b",
   "chip": "RX 470",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 207.16,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 40.11
  },
  {
   "backend": "Vulkan",
   "build": "b8372ee",
   "chip": "Tesla M40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 92.48,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 39.35
  },
  {
   "backend": "Vulkan",
   "build": "de4c07f",
   "chip": "RX 580",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 258.03,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 39.32
  },
  {
   "backend": "Vulkan",
   "build": "e583f3b4f",
   "chip": "RX 570",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 226.11,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 37.44
  },
  {
   "backend": "Vulkan",
   "build": "860a9e4",
   "chip": "Radeon Pro W5500",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 315.39,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 36.82
  },
  {
   "backend": "Vulkan",
   "build": "3b15924",
   "chip": "RX 480",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 248.66,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 34.71
  },
  {
   "backend": "Vulkan",
   "build": "dbb852b",
   "chip": "Apple M2 Ultra",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 205.98,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 34.34
  },
  {
   "backend": "Vulkan",
   "build": "860a9e4",
   "chip": "GTX 980",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 186.24,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 33.9
  },
  {
   "backend": "Vulkan",
   "build": "23bc779",
   "chip": "P106-100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 183.78,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 29.77
  },
  {
   "backend": "Vulkan",
   "build": "45363632c",
   "chip": "AMD FirePro W8100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 155.22,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 29.52
  },
  {
   "backend": "Vulkan",
   "build": "24d2ee0",
   "chip": "Tesla P4",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 265.54,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 28.03
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "RX 6500 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 255.25,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 27.81
  },
  {
   "backend": "Vulkan",
   "build": "b9ab0a4",
   "chip": "Apple M3",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 263.7,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 26.39
  },
  {
   "backend": "Vulkan",
   "build": "914a82d",
   "chip": "AMD FirePro S10000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 94.78,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 25.32
  },
  {
   "backend": "Vulkan",
   "build": "481c65f",
   "chip": "Intel Core Ultra 300 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1380.07,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 24.45
  },
  {
   "backend": "Vulkan",
   "build": "4a89937",
   "chip": "Intel Core Ultra 200 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 864.99,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 24.37
  },
  {
   "backend": "Vulkan",
   "build": "63f8fe0",
   "chip": "Quadro P2000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 169.55,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 23.05
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "AMD Ryzen AI 300 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 479.07,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 22.41
  },
  {
   "backend": "Vulkan",
   "build": "ee09828",
   "chip": "AMD Ryzen 6000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 240.89,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 21.26
  },
  {
   "backend": "Vulkan",
   "build": "1fe0029",
   "chip": "Apple M2 Pro",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 62.7,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 20.95
  },
  {
   "backend": "Vulkan",
   "build": "2f0c2db43",
   "chip": "GTX 1050 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 136.42,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 20.96
  },
  {
   "backend": "Vulkan",
   "build": "ebce03e",
   "chip": "AMD Ryzen 7000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 281.62,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 19.91
  },
  {
   "backend": "Vulkan",
   "build": "1e9d771",
   "chip": "AMD Ryzen 8000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 267.27,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 19.81
  },
  {
   "backend": "Vulkan",
   "build": "53ff6b9",
   "chip": "AMD Ryzen Z1 Extreme",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 199.36,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 18.77
  },
  {
   "backend": "Vulkan",
   "build": "cb26014",
   "chip": "AMD Ryzen AI 400 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 131.33,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 17.66
  },
  {
   "backend": "Vulkan",
   "build": "48d22e2",
   "chip": "AMD Steam Deck",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 144.31,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 17.52
  },
  {
   "backend": "Vulkan",
   "build": "d3bd7193",
   "chip": "AMD FirePro D700",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 69.95,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 16.62
  },
  {
   "backend": "Vulkan",
   "build": "860a9e4",
   "chip": "Radeon Pro WX 4100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 78.79,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 16.05
  },
  {
   "backend": "Vulkan",
   "build": "8c0d6bb",
   "chip": "Apple M2",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 50.79,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 13.5
  },
  {
   "backend": "Vulkan",
   "build": "2370665",
   "chip": "Apple M1",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 38.29,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 12.47
  },
  {
   "backend": "Vulkan",
   "build": "d84635b",
   "chip": "AMD Ryzen 5000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 90.55,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 10.98
  },
  {
   "backend": "Vulkan",
   "build": "abb9f3c42",
   "chip": "Intel Core 1100 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 187.2,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 10.39
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "RX 550",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 52.66,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 10.2
  },
  {
   "backend": "Vulkan",
   "build": "4b385bf",
   "chip": "AMD Ryzen 4000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 103.87,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 9.63
  },
  {
   "backend": "Vulkan",
   "build": "5d46bab",
   "chip": "Tesla K80",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 89.46,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 9.39
  },
  {
   "backend": "Vulkan",
   "build": "eec1e33",
   "chip": "Tesla K40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 64.37,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 9.3
  },
  {
   "backend": "Vulkan",
   "build": "b9ab0a4",
   "chip": "MediaTek Dimensity 9400",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 38.36,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 8.92
  },
  {
   "backend": "Vulkan",
   "build": "1d72c84",
   "chip": "Intel Core Ultra 100 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 185.51,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 8.21
  },
  {
   "backend": "Vulkan",
   "build": "1fe0029",
   "chip": "AMD Ryzen 3000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 48.63,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 8.49
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "Intel Core 1000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 25.58,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 4.25
  },
  {
   "backend": "Vulkan",
   "build": "c4df49a",
   "chip": "Intel Core 8000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 25.43,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 3.35
  },
  {
   "backend": "Vulkan",
   "build": "4f63cd7",
   "chip": "Intel N150",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 28.84,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 2.93
  },
  {
   "backend": "Vulkan",
   "build": "ca71fb9",
   "chip": "RTX 5090",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 11796.38,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 273.68
  },
  {
   "backend": "Vulkan",
   "build": "f6b533d89",
   "chip": "RTX 5080",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 8054.59,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 192.17
  },
  {
   "backend": "Vulkan",
   "build": "304665f",
   "chip": "RX 7900 XTX",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3889.54,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 190.92
  },
  {
   "backend": "Vulkan",
   "build": "4ae88d0",
   "chip": "RTX 4090",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 10830.41,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 190.1
  },
  {
   "backend": "Vulkan",
   "build": "0d0764d",
   "chip": "RTX 5070 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 7567.94,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 176.18
  },
  {
   "backend": "Vulkan",
   "build": "d05fe1d",
   "chip": "RTX 3090",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5200.68,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 171.61
  },
  {
   "backend": "Vulkan",
   "build": "2257758",
   "chip": "A100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 7064.4,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 170.56
  },
  {
   "backend": "Vulkan",
   "build": "4bc272f",
   "chip": "RTX 5090 Mobile",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 6011.18,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 154.31
  },
  {
   "backend": "Vulkan",
   "build": "dd1ea5243",
   "chip": "AI Pro R9700",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5907.66,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 152.7
  },
  {
   "backend": "Vulkan",
   "build": "81086cd",
   "chip": "RTX 4080 Super",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 8007.37,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 150.2
  },
  {
   "backend": "Vulkan",
   "build": "7c7d6ce",
   "chip": "RTX 3080",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4913.83,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 145.74
  },
  {
   "backend": "Vulkan",
   "build": "7d77f07",
   "chip": "Tesla V100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1411.25,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 142.13
  },
  {
   "backend": "Vulkan",
   "build": "4ae88d0",
   "chip": "RTX A5000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4071.22,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 140.43
  },
  {
   "backend": "Vulkan",
   "build": "4ae88d0",
   "chip": "RTX 4070 Ti Super",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 6801.18,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 135.81
  },
  {
   "backend": "Vulkan",
   "build": "e9fd8dc",
   "chip": "RX 9070 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5048.07,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 131.54
  },
  {
   "backend": "Vulkan",
   "build": "4fdbc1e",
   "chip": "RX 7800 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2197.05,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 124.86
  },
  {
   "backend": "Vulkan",
   "build": "71e74a3",
   "chip": "RX 7900 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2701.13,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 120.62
  },
  {
   "backend": "Vulkan",
   "build": "21c17b5",
   "chip": "RX 9070",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2859.98,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 119.51
  },
  {
   "backend": "Vulkan",
   "build": "3af34b9",
   "chip": "Instinct MI50",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1127.37,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 117.94
  },
  {
   "backend": "Vulkan",
   "build": "583cb83",
   "chip": "A30",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3158.25,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 117.43
  },
  {
   "backend": "Vulkan",
   "build": "b863507",
   "chip": "Intel Arc Pro B70",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3150.55,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 114.19
  },
  {
   "backend": "Vulkan",
   "build": "516a4ca",
   "chip": "RTX 4070 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5515.26,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 113.52
  },
  {
   "backend": "Vulkan",
   "build": "c945aaa",
   "chip": "RTX 4070 Super",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 5198.91,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 112.0
  },
  {
   "backend": "Vulkan",
   "build": "77d6ae4",
   "chip": "Radeon VII",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1007.19,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 110.86
  },
  {
   "backend": "Vulkan",
   "build": "4b2a477",
   "chip": "RX 7900 GRE",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2251.37,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 110.5
  },
  {
   "backend": "Vulkan",
   "build": "e56abd2",
   "chip": "Titan V",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 792.74,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 109.21
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "Radeon Pro VII",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 783.94,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 108.45
  },
  {
   "backend": "Vulkan",
   "build": "f1cee99",
   "chip": "RX 6800",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1597.95,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 106.49
  },
  {
   "backend": "Vulkan",
   "build": "a972fae",
   "chip": "RX 6900 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1761.93,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 106.15
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "RTX 2080 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1936.25,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 100.99
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "RX 6800 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1704.79,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 100.5
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "Radeon Pro W6800X Duo",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 795.28,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 100.08
  },
  {
   "backend": "Vulkan",
   "build": "89f10ba",
   "chip": "RTX 5060 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3912.65,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 97.01
  },
  {
   "backend": "Vulkan",
   "build": "9a483999",
   "chip": "RTX 4070",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 4293.57,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 91.49
  },
  {
   "backend": "Vulkan",
   "build": "b42c7fa5b",
   "chip": "RTX 3060 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2644.95,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 90.35
  },
  {
   "backend": "Vulkan",
   "build": "228f34c",
   "chip": "RX 6750 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 997.05,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 82.29
  },
  {
   "backend": "Vulkan",
   "build": "6d75883",
   "chip": "RX 6700 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1010.9,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 81.86
  },
  {
   "backend": "Vulkan",
   "build": "92c0b38",
   "chip": "RTX 3060",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2012.88,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 80.59
  },
  {
   "backend": "Vulkan",
   "build": "03d4698",
   "chip": "Radeon Pro V620",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1556.31,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 79.24
  },
  {
   "backend": "Vulkan",
   "build": "f5245b5",
   "chip": "RTX A4000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2482.74,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 76.07
  },
  {
   "backend": "Vulkan",
   "build": "7f76692",
   "chip": "Tesla T10",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1840.14,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 76.05
  },
  {
   "backend": "Vulkan",
   "build": "4fdbc1e",
   "chip": "RX 5700 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 538.31,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 74.43
  },
  {
   "backend": "Vulkan",
   "build": "f1cee99",
   "chip": "RX 9060 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3228.15,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 74.01
  },
  {
   "backend": "Vulkan",
   "build": "7f76692",
   "chip": "Intel Arc B580",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 419.49,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 72.0
  },
  {
   "backend": "Vulkan",
   "build": "1ece0cb6",
   "chip": "Apple M4 Max",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 557.46,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 71.79
  },
  {
   "backend": "Vulkan",
   "build": "ac2557c",
   "chip": "GTX 1080 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 647.81,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 71.58
  },
  {
   "backend": "Vulkan",
   "build": "23bc779",
   "chip": "Radeon Pro W5700",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 446.98,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 71.3
  },
  {
   "backend": "Vulkan",
   "build": "516a4ca",
   "chip": "Intel Arc Pro B60",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 274.76,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 70.54
  },
  {
   "backend": "Vulkan",
   "build": "5dcb711",
   "chip": "RX Vega 64",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 584.98,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 67.7
  },
  {
   "backend": "Vulkan",
   "build": "eec1e33",
   "chip": "Tesla P100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 685.51,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 66.48
  },
  {
   "backend": "Vulkan",
   "build": "ac2557c",
   "chip": "P102-100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 574.56,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 66.48
  },
  {
   "backend": "Vulkan",
   "build": "5886f4f54",
   "chip": "AMD BC-250",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 356.87,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 63.14
  },
  {
   "backend": "Vulkan",
   "build": "ceff6bb",
   "chip": "RTX 3070 Mobile",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1832.07,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 62.92
  },
  {
   "backend": "Vulkan",
   "build": "dbb852b",
   "chip": "RX 6650 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1088.9,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 64.53
  },
  {
   "backend": "Vulkan",
   "build": "91f8c9c",
   "chip": "RX 7600",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1353.47,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 60.88
  },
  {
   "backend": "Vulkan",
   "build": "657e011",
   "chip": "GB10",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 3321.91,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 60.39
  },
  {
   "backend": "Vulkan",
   "build": "a5c07dc",
   "chip": "RTX 4060 Mobile",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 2358.03,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 60.01
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "Tesla P40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 484.37,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 59.22
  },
  {
   "backend": "Vulkan",
   "build": "b43556e",
   "chip": "GTX 1660 Ti Mobile",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 514.34,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 57.3
  },
  {
   "backend": "Vulkan",
   "build": "01d8eaa",
   "chip": "RX 7600 XT",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1024.38,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 56.11
  },
  {
   "backend": "Vulkan",
   "build": "eec1e33",
   "chip": "AMD FirePro S9300 x2",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 243.33,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 55.64
  },
  {
   "backend": "Vulkan",
   "build": "fb4b273",
   "chip": "AMD Ryzen AI Max+ 300 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1704.3,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 54.38
  },
  {
   "backend": "Vulkan",
   "build": "86a9c79",
   "chip": "RX 6600",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 804.64,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 53.17
  },
  {
   "backend": "Vulkan",
   "build": "a69d54f",
   "chip": "Intel Arc A770",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1119.68,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 53.07
  },
  {
   "backend": "Vulkan",
   "build": "92c0b38",
   "chip": "RX Vega 56",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 428.54,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 52.66
  },
  {
   "backend": "Vulkan",
   "build": "7f76692",
   "chip": "Intel Arc B570",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 288.51,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 50.49
  },
  {
   "backend": "Vulkan",
   "build": "8e6f8bc8",
   "chip": "RX 6800M",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 784.16,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 49.06
  },
  {
   "backend": "Vulkan",
   "build": "b1afcab",
   "chip": "RTX A2000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1361.85,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 45.69
  },
  {
   "backend": "Vulkan",
   "build": "eeee367de",
   "chip": "Intel Arc A770M",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 384.74,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 45.68
  },
  {
   "backend": "Vulkan",
   "build": "eec1e33",
   "chip": "P104-100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 325.3,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 48.64
  },
  {
   "backend": "Vulkan",
   "build": "aa46bda",
   "chip": "Intel Arc Pro B50",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1122.53,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 47.69
  },
  {
   "backend": "Vulkan",
   "build": "9da3dcd",
   "chip": "Radeon Pro V340",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 360.23,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 47.54
  },
  {
   "backend": "Vulkan",
   "build": "c1b1876",
   "chip": "Intel Arc A750",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 303.37,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 43.96
  },
  {
   "backend": "Vulkan",
   "build": "860a9e4",
   "chip": "GTX 1070 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 292.85,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 43.42
  },
  {
   "backend": "Vulkan",
   "build": "32e41fa5b",
   "chip": "RX 470",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 211.25,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 43.35
  },
  {
   "backend": "Vulkan",
   "build": "360d653",
   "chip": "GTX 1070",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 330.84,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 43.33
  },
  {
   "backend": "Vulkan",
   "build": "b8372ee",
   "chip": "Tesla M40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 93.35,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 41.68
  },
  {
   "backend": "Vulkan",
   "build": "e583f3b4f",
   "chip": "RX 570",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 229.23,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 40.18
  },
  {
   "backend": "Vulkan",
   "build": "0bcb40b",
   "chip": "RX 480",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 194.52,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 37.23
  },
  {
   "backend": "Vulkan",
   "build": "dbb852b",
   "chip": "Apple M2 Ultra",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 198.83,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 36.07
  },
  {
   "backend": "Vulkan",
   "build": "860a9e4",
   "chip": "GTX 980",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 180.97,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 34.16
  },
  {
   "backend": "Vulkan",
   "build": "23bc779",
   "chip": "P106-100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 183.4,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 30.79
  },
  {
   "backend": "Vulkan",
   "build": "45363632c",
   "chip": "AMD FirePro W8100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 140.52,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 29.28
  },
  {
   "backend": "Vulkan",
   "build": "24d2ee0",
   "chip": "Tesla P4",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 287.14,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 28.37
  },
  {
   "backend": "Vulkan",
   "build": "4a89937",
   "chip": "Intel Core Ultra 200 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 867.27,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 25.12
  },
  {
   "backend": "Vulkan",
   "build": "481c65f",
   "chip": "Intel Core Ultra 300 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 1332.41,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 23.93
  },
  {
   "backend": "Vulkan",
   "build": "63f8fe0",
   "chip": "Quadro P2000",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 181.71,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 23.77
  },
  {
   "backend": "Vulkan",
   "build": "",
   "chip": "AMD Ryzen AI 300 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 532.59,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 23.93
  },
  {
   "backend": "Vulkan",
   "build": "1e9d771",
   "chip": "AMD Ryzen 8000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 311.02,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 21.28
  },
  {
   "backend": "Vulkan",
   "build": "ee09828",
   "chip": "AMD Ryzen 6000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 277.91,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 21.15
  },
  {
   "backend": "Vulkan",
   "build": "1fe0029",
   "chip": "Apple M2 Pro",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 58.86,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 20.97
  },
  {
   "backend": "Vulkan",
   "build": "835b2b9",
   "chip": "AMD Ryzen 7000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 312.85,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 20.09
  },
  {
   "backend": "Vulkan",
   "build": "2f0c2db43",
   "chip": "GTX 1050 Ti",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 127.54,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 20.08
  },
  {
   "backend": "Vulkan",
   "build": "cb26014",
   "chip": "AMD Ryzen AI 400 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 150.12,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 19.03
  },
  {
   "backend": "Vulkan",
   "build": "48d22e2",
   "chip": "AMD Steam Deck",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 164.34,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 18.4
  },
  {
   "backend": "Vulkan",
   "build": "860a9e4",
   "chip": "Radeon Pro WX 4100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 75.59,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 16.56
  },
  {
   "backend": "Vulkan",
   "build": "2370665",
   "chip": "Apple M1",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 35.93,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 12.85
  },
  {
   "backend": "Vulkan",
   "build": "8c0d6bb",
   "chip": "Apple M2",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 46.81,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 12.25
  },
  {
   "backend": "Vulkan",
   "build": "5d195f17b",
   "chip": "AMD Ryzen 5000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 79.06,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 10.75
  },
  {
   "backend": "Vulkan",
   "build": "abb9f3c42",
   "chip": "Intel Core 1100 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 174.77,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 10.58
  },
  {
   "backend": "Vulkan",
   "build": "eec1e33",
   "chip": "Tesla K40",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 64.37,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 9.92
  },
  {
   "backend": "Vulkan",
   "build": "4b385bf",
   "chip": "AMD Ryzen 4000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 113.32,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 9.87
  },
  {
   "backend": "Vulkan",
   "build": "5d46bab",
   "chip": "Tesla K80",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 88.26,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 9.49
  },
  {
   "backend": "Vulkan",
   "build": "1fe0029",
   "chip": "AMD Ryzen 5 3000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 47.41,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 8.47
  },
  {
   "backend": "Vulkan",
   "build": "2e89f76",
   "chip": "Intel Core Ultra 100 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 77.66,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 7.75
  },
  {
   "backend": "Vulkan",
   "build": "c4df49a",
   "chip": "Intel Core 8000 Series",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 25.55,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 3.35
  },
  {
   "backend": "Vulkan",
   "build": "4f63cd7",
   "chip": "Intel N150",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 25.59,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 10879",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "write_tps": 2.91
  },
  {
   "backend": "CPU",
   "build": "a27152b6",
   "chip": "Qualcomm Snapdragon X Plus X1P-64-100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 58.72,
   "quant": "Q8_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 8273",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/8273",
   "write_tps": 13.54
  },
  {
   "backend": "CPU",
   "build": "a27152b6",
   "chip": "Qualcomm Snapdragon X Plus X1P-64-100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 58.59,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 8273",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/8273",
   "write_tps": 18.23
  },
  {
   "backend": "CPU",
   "build": "a27152b6",
   "chip": "Qualcomm Snapdragon X Plus X1P-64-100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 62.46,
   "quant": "Q8_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 8273",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/8273",
   "write_tps": 9.58
  },
  {
   "backend": "CPU",
   "build": "a27152b6",
   "chip": "Qualcomm Snapdragon X Plus X1P-64-100",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 61.93,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 8273",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/8273",
   "write_tps": 13.74
  },
  {
   "backend": "CPU",
   "build": "cddae488",
   "chip": "Qualcomm Snapdragon X Elite (12-core)",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 63.51,
   "quant": "Q8_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 8273",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/8273",
   "write_tps": 12.65
  },
  {
   "backend": "CPU",
   "build": "cddae488",
   "chip": "Qualcomm Snapdragon X Elite (12-core)",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 66.63,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 8273",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/8273",
   "write_tps": 20.72
  },
  {
   "backend": "CPU",
   "build": "fa42aa6d",
   "chip": "Qualcomm Snapdragon X Elite (12-core)",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 63.05,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 8273",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/8273",
   "write_tps": 19.83
  },
  {
   "backend": "CPU",
   "build": "d5fe4e81",
   "chip": "Qualcomm Snapdragon X Elite (12-core)",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 166.4,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 8273",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/8273",
   "write_tps": 19.91
  },
  {
   "backend": "OpenCL",
   "build": "d5fe4e81",
   "chip": "Qualcomm Adreno X1-85 GPU (Snapdragon X Elite)",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 101.53,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 8273",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/8273",
   "write_tps": 17.61
  },
  {
   "backend": "OpenCL",
   "build": "d5fe4e81",
   "chip": "Qualcomm Adreno X1-85 GPU (Snapdragon X Elite)",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 139.62,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 8273",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/8273",
   "write_tps": 10.86
  },
  {
   "backend": "CPU",
   "build": "",
   "chip": "Qualcomm Snapdragon X Elite (12-core)",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 177.71,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 8336",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/8336",
   "write_tps": 24.81
  },
  {
   "backend": "CPU",
   "build": "",
   "chip": "Qualcomm Snapdragon X Elite (12-core)",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 177.35,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 8336",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/8336",
   "write_tps": 22.55
  },
  {
   "backend": "OpenCL",
   "build": "a76c56fa",
   "chip": "Qualcomm Adreno X1-85 GPU (Snapdragon X Elite)",
   "gpu_cores": null,
   "model": "Llama 2 7B",
   "prompt_tps": 100.65,
   "quant": "Q4_0",
   "retrieved": "2026-10-20",
   "source": "llama.cpp discussion 8336",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/8336",
   "write_tps": 17.95
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4090",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 7594.49,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1704",
   "write_tps": 119.61
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX 6000 Ada Generation",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 6807.82,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/673",
   "write_tps": 120.98
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4090",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 6696.53,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/77",
   "write_tps": 91.03
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 6345.71,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/937",
   "write_tps": 85.77
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA H100 PCIe",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 5202.16,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/69",
   "write_tps": 119.97
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5090",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 6297.37,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/155",
   "write_tps": 66.33
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5090 Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 4462.93,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/4016",
   "write_tps": 105.09
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4080",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 4791.73,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/15",
   "write_tps": 88.34
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4080 SUPER",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 4978.68,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2949",
   "write_tps": 78.69
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3090 Ti",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 4023.78,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/127",
   "write_tps": 109.78
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA A100-SXM4-80GB",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3570.51,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/61",
   "write_tps": 110.1
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 5543.92,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1344",
   "write_tps": 46.21
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX 6000 Ada Generation",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 5487.5,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/86",
   "write_tps": 51.29
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5090 Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3829.46,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3553",
   "write_tps": 90.56
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA L40S",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 5327.55,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/84",
   "write_tps": 45.91
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3090",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3549.56,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1",
   "write_tps": 96.28
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3080 Ti",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3514.63,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/120",
   "write_tps": 90.49
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA L40",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 5064.74,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/82",
   "write_tps": 45.15
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX A6000",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3447.07,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/50",
   "write_tps": 90.48
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4080 SUPER",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 4461.13,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/102",
   "write_tps": 54.44
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA H100 80GB HBM3",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 4686.8,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/87",
   "write_tps": 42.41
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5080",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 4427.98,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/489",
   "write_tps": 44.91
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5070 Ti",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3656.83,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/160",
   "write_tps": 65.54
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4070 Ti",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3768.73,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/138",
   "write_tps": 64.23
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA A40",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3270.17,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/49",
   "write_tps": 80.58
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4070",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3192.12,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/147",
   "write_tps": 76.35
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3080",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3030.93,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2948",
   "write_tps": 81.65
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX A5000",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2890.51,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/26",
   "write_tps": 89.05
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3080",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3018.71,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/132",
   "write_tps": 80.58
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA H100 NVL",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 4298.75,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/78",
   "write_tps": 37.06
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4070 Ti SUPER",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3561.02,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/136",
   "write_tps": 53.94
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4070 SUPER",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3335.66,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/315",
   "write_tps": 53.39
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5070 Ti Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2874.33,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3548",
   "write_tps": 65.41
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3070 Ti",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2509.44,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/122",
   "write_tps": 83.17
  },
  {
   "backend": "",
   "build": "",
   "chip": "Tesla V100-SXM2-16GB",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2256.05,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/775",
   "write_tps": 90.82
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX A4500",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2498.48,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/51",
   "write_tps": 77.91
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5070",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2927.06,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/168",
   "write_tps": 55.92
  },
  {
   "backend": "",
   "build": "",
   "chip": "Tesla V100-SXM2-32GB",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2176.54,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/920",
   "write_tps": 88.17
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA TITAN RTX",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2226.04,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3817",
   "write_tps": 71.22
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5080",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 3662.38,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1119",
   "write_tps": 26.11
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX 4000 Ada Generation",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2608.05,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/76",
   "write_tps": 53.54
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 2080 Ti",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2026.98,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/112",
   "write_tps": 72.91
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5060 Ti",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2364.57,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/860",
   "write_tps": 59.01
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA TITAN V",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1822.65,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/107",
   "write_tps": 73.34
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5070 Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2030.2,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/4230",
   "write_tps": 57.68
  },
  {
   "backend": "",
   "build": "",
   "chip": "Quadro RTX 6000",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1891.12,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3523",
   "write_tps": 64.86
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA A100 80GB PCIe",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2600.69,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/64",
   "write_tps": 37.1
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX 3500 Ada Generation Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2444.97,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/632",
   "write_tps": 41.79
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5070 Ti",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2574.58,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1847",
   "write_tps": 34.69
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3070",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2005.7,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/126",
   "write_tps": 59.6
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4060 Ti",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2233.12,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/7",
   "write_tps": 48.32
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3070 Ti Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1909.66,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/741",
   "write_tps": 62.35
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3080 Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2052.3,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2029",
   "write_tps": 53.87
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX A4000",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1930.9,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/44",
   "write_tps": 56.5
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3060 Ti",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1839.18,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/121",
   "write_tps": 60.16
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5060 Ti",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2313.58,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2298",
   "write_tps": 51.2
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3080 Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1884.99,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/894",
   "write_tps": 52.29
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4070 Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2130.57,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1537",
   "write_tps": 40.88
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 2080 SUPER",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1693.29,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2011",
   "write_tps": 59.8
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA L4",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2056.48,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/63",
   "write_tps": 42.87
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX 4000 SFF Ada Generation",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2003.96,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/157",
   "write_tps": 44.37
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4060 Ti",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2086.64,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/328",
   "write_tps": 38.29
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA A10",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2085.03,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/363",
   "write_tps": 36.5
  },
  {
   "backend": "",
   "build": "",
   "chip": "Quadro RTX 8000",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1677.26,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/569",
   "write_tps": 74.02
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 2080",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1545.77,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1476",
   "write_tps": 58.43
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GB10",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 2009.91,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2944",
   "write_tps": 34.7
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GB10",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1986.31,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1771",
   "write_tps": 33.63
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 2070 SUPER",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1443.97,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/101",
   "write_tps": 57.59
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3060",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1482.57,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/43",
   "write_tps": 51.28
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX A5000 Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1497.11,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1024",
   "write_tps": 45.47
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4060",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1544.18,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/137",
   "write_tps": 38.3
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Ultra 24P+8E+80GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1109.46,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/404",
   "write_tps": 62.68
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 2070",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1251.93,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1727",
   "write_tps": 52.39
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX PRO 2000 Blackwell Generation Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1453.06,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1487",
   "write_tps": 40.4
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Ultra 24P+8E+80GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1062.18,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1359",
   "write_tps": 63.3
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 2060 SUPER",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1205.22,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/111",
   "write_tps": 49.92
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX 2000 Ada Generation",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1356.95,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/62",
   "write_tps": 37.55
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3060",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1339.87,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/4053",
   "write_tps": 36.97
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3070 Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1175.73,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3572",
   "write_tps": 41.88
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3060 Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1252.83,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/767",
   "write_tps": 40.76
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4060 Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1365.05,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/529",
   "write_tps": 35.97
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon RX 6900 XT",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1082.48,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/319",
   "write_tps": 42.42
  },
  {
   "backend": "",
   "build": "",
   "chip": "Quadro RTX 4000",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 992.39,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1463",
   "write_tps": 40.0
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4050 Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1200.99,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/598",
   "write_tps": 31.36
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Max 6P+12E+40GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 718.42,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/4247",
   "write_tps": 71.25
  },
  {
   "backend": "",
   "build": "",
   "chip": "Radeon RX 7900 XTX",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 885.1,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/880",
   "write_tps": 45.33
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 2060",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 960.07,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/665",
   "write_tps": 38.08
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX 2000 Ada Generation Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1150.82,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3256",
   "write_tps": 29.51
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX A2000 12GB",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 976.19,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/514",
   "write_tps": 35.73
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Ultra 16P+8E+60GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 705.74,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2309",
   "write_tps": 59.4
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3050",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 979.49,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2294",
   "write_tps": 34.47
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3050 OEM",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 964.79,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3339",
   "write_tps": 34.96
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon RX 6800 XT",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 950.57,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/475",
   "write_tps": 37.64
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Ultra 16P+4E+64GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 667.57,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1087",
   "write_tps": 54.31
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Max 6P+12E+32GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 637.38,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3105",
   "write_tps": 62.19
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon RX 7900 XTX",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 864.54,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/491",
   "write_tps": 39.25
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Max 12P+4E+40GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 656.49,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1188",
   "write_tps": 54.47
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Max 12P+4E+40GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 631.76,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/6",
   "write_tps": 52.85
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX 1000 Ada Generation Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 954.71,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3883",
   "write_tps": 24.18
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Max 12P+4E+40GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 565.91,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/776",
   "write_tps": 47.94
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX A2000 8GB Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 790.31,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/731",
   "write_tps": 27.52
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Max 12P+4E+40GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 582.55,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/750",
   "write_tps": 45.76
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Ultra 16P+4E+48GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 534.35,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/306",
   "write_tps": 48.89
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon RX 6800",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 737.27,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/494",
   "write_tps": 30.88
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3050 6GB Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 789.78,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1998",
   "write_tps": 24.2
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX PRO 1000 Blackwell Generation Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 906.23,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2077",
   "write_tps": 19.85
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Max 10P+4E+32GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 540.04,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/162",
   "write_tps": 45.5
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX A2000",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 758.3,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1425",
   "write_tps": 28.73
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3050",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 778.4,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/977",
   "write_tps": 25.92
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Max 8P+4E+38GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 484.06,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/515",
   "write_tps": 46.41
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon RX 7800 XT",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 524.49,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1003",
   "write_tps": 38.89
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Max 8P+4E+38GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 473.66,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1291",
   "write_tps": 44.66
  },
  {
   "backend": "",
   "build": "",
   "chip": "Quadro RTX 3000",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 606.47,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3949",
   "write_tps": 29.47
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce GTX 1080",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 652.37,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/110",
   "write_tps": 20.23
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Max 10P+4E+30GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 456.81,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/789",
   "write_tps": 37.74
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Pro 6P+12E+20GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 411.48,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3180",
   "write_tps": 44.9
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Max 10P+4E+30GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 443.5,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/805",
   "write_tps": 37.48
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Pro 6P+12E+20GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 410.4,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3226",
   "write_tps": 44.53
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon RX 6600 XT",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 604.46,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/624",
   "write_tps": 21.5
  },
  {
   "backend": "",
   "build": "",
   "chip": "Tesla P100-SXM2-16GB",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 427.13,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/100",
   "write_tps": 33.46
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon RX 6650 XT",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 535.6,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/12",
   "write_tps": 25.25
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Max 8P+2E+32GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 372.19,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/364",
   "write_tps": 36.66
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon RX 6950 XT",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 1023.72,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1664",
   "write_tps": 6.14
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Pro 5P+10E+16GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 340.58,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3323",
   "write_tps": 43.74
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon RX 7800 XT",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 402.68,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/607",
   "write_tps": 36.7
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Pro 5P+10E+16GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 336.23,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3531",
   "write_tps": 43.16
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Max 8P+2E+32GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 365.64,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/597",
   "write_tps": 35.85
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Max 8P+4E+30GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 362.07,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1079",
   "write_tps": 36.23
  },
  {
   "backend": "",
   "build": "",
   "chip": "Tesla P100-PCIE-16GB",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 371.51,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3387",
   "write_tps": 31.07
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Pro 10P+4E+20GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 361.47,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/777",
   "write_tps": 32.68
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Pro 10P+4E+20GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 359.56,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/561",
   "write_tps": 32.53
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon RX 6700 XT",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 596.91,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/605",
   "write_tps": 16.37
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Pro 10P+4E+20GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 349.41,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/369",
   "write_tps": 32.9
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Max 12P+4E+40GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 377.05,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1469",
   "write_tps": 25.45
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Max 8P+2E+24GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 306.19,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/688",
   "write_tps": 32.12
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Pro 8P+4E+16GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 302.0,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/550",
   "write_tps": 30.59
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Pro 8P+4E+16GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 292.9,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/380",
   "write_tps": 29.72
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Pro 8P+4E+19GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 261.76,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/470",
   "write_tps": 26.31
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Pro 6P+6E+18GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 283.81,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/428",
   "write_tps": 22.06
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M6 2P+4E+12GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 263.15,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/4201",
   "write_tps": 25.24
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Pro 6P+6E+18GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 279.21,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/401",
   "write_tps": 20.81
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon RX 5700 XT",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 256.38,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/804",
   "write_tps": 25.35
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon RX 6600",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 418.8,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1078",
   "write_tps": 10.24
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Pro 6P+4E+16GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 224.56,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1220",
   "write_tps": 24.33
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Pro 6P+4E+16GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 216.28,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1298",
   "write_tps": 24.14
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Pro 5P+6E+14GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 223.44,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/862",
   "write_tps": 21.46
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Pro 5P+6E+14GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 217.77,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/263",
   "write_tps": 21.12
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Pro 8P+2E+16GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 204.54,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/929",
   "write_tps": 21.88
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 4P+6E+10GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 202.38,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2912",
   "write_tps": 21.62
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Pro 8P+2E+16GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 200.28,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/4",
   "write_tps": 21.7
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 4P+6E+10GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 188.54,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1716",
   "write_tps": 20.24
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 4P+6E+10GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 176.95,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3811",
   "write_tps": 20.01
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Pro 6P+2E+14GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 176.97,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/412",
   "write_tps": 20.14
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Pro 6P+2E+14GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 173.08,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/418",
   "write_tps": 19.97
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon RX 7600",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 317.11,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1556",
   "write_tps": 5.46
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 4P+4E+10GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 168.08,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/847",
   "write_tps": 19.43
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 4P+6E+10GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 175.22,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/32",
   "write_tps": 16.9
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 4P+6E+10GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 166.15,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/546",
   "write_tps": 16.81
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon™ RX 7700S",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 289.27,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3710",
   "write_tps": 4.56
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 4P+4E+8GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 143.61,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/245",
   "write_tps": 17.32
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 4P+6E+10GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 147.89,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1085",
   "write_tps": 15.44
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 4P+4E+10GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 142.38,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/202",
   "write_tps": 14.74
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 4P+4E+10GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 141.26,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/499",
   "write_tps": 14.65
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 4P+6E+8GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 134.03,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1394",
   "write_tps": 15.27
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 4P+4E+10GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 138.71,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/414",
   "write_tps": 13.5
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 4P+4E+8GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 134.01,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/17",
   "write_tps": 14.58
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3050 Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 279.67,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1642",
   "write_tps": 4.75
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 4P+4E+8GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 124.94,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3123",
   "write_tps": 13.7
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple A18 Pro 2P+4E+5GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 115.56,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3128",
   "write_tps": 15.83
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon 780M Graphics",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 186.03,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3235",
   "write_tps": 5.63
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3050 Ti Laptop GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 290.0,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/366",
   "write_tps": 4.3
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 4P+4E+8GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 113.96,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1939",
   "write_tps": 12.91
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 4P+4E+7GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 108.72,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1281",
   "write_tps": 13.39
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce GTX 1660 SUPER",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 129.2,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1595",
   "write_tps": 9.59
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce GTX 1660",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 128.69,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/727",
   "write_tps": 9.01
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 4P+4E+10GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 109.9,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/562",
   "write_tps": 10.23
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 4P+4E+8GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 92.66,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/432",
   "write_tps": 9.3
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 4P+4E+7GPU",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 83.94,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/223",
   "write_tps": 9.39
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon(TM) 780M",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 111.7,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1578",
   "write_tps": 3.8
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon 780M Graphics",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 109.73,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3179",
   "write_tps": 3.5
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon 780M Graphics",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 103.43,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3862",
   "write_tps": 3.29
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon 780M Graphics",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 93.86,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1965",
   "write_tps": 3.31
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA T1000 8GB",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 74.5,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/4232",
   "write_tps": 5.23
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce GTX 1650 Ti",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 65.63,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/374",
   "write_tps": 2.8
  },
  {
   "backend": "",
   "build": "",
   "chip": "Quadro P620",
   "gpu_cores": null,
   "model": "Meta Llama 3.1 8B Instruct",
   "prompt_tps": 45.39,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3526",
   "write_tps": 1.31
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 4P+4E+10GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 4916.16,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/414",
   "write_tps": 783.94
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2622.65,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1344",
   "write_tps": 23.71
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5090 Laptop GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1549.65,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3553",
   "write_tps": 40.47
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5070 Ti Laptop GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1473.59,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3548",
   "write_tps": 36.29
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX 3500 Ada Generation Laptop GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1297.99,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/632",
   "write_tps": 25.06
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3080 Laptop GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 858.4,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/894",
   "write_tps": 28.21
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX A5000 Laptop GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 839.43,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1024",
   "write_tps": 27.58
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX 5000 Ada Generation Laptop GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 951.71,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1357",
   "write_tps": 16.39
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Ultra 24P+8E+80GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 580.03,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/404",
   "write_tps": 35.82
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Ultra 24P+8E+80GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 568.25,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1359",
   "write_tps": 36.71
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Ultra 16P+8E+76GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 470.53,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1222",
   "write_tps": 36.63
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Ultra 16P+8E+76GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 469.48,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3670",
   "write_tps": 36.53
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Max 6P+12E+40GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 424.13,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/4247",
   "write_tps": 42.1
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Ultra 20P+8E+60GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 444.62,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/479",
   "write_tps": 34.44
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Ultra 16P+8E+60GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 381.13,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2309",
   "write_tps": 34.21
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Ultra 16P+8E+60GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 380.08,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/617",
   "write_tps": 33.92
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Max 6P+12E+40GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 359.53,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3148",
   "write_tps": 36.87
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Ultra 16P+4E+64GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 371.72,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1087",
   "write_tps": 32.42
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Max 6P+12E+32GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 343.13,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3105",
   "write_tps": 34.15
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Max 12P+4E+40GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 335.92,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1188",
   "write_tps": 30.23
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Max 12P+4E+40GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 328.43,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/6",
   "write_tps": 28.95
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Ultra 16P+4E+48GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 289.9,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/306",
   "write_tps": 27.77
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Max 12P+4E+40GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 302.37,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/750",
   "write_tps": 25.49
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Max 12P+4E+40GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 287.77,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/776",
   "write_tps": 25.62
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Max 10P+4E+32GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 284.61,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/162",
   "write_tps": 25.56
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Max 12P+4E+40GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 286.99,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1114",
   "write_tps": 23.64
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Max 8P+4E+38GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 252.47,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/515",
   "write_tps": 25.23
  },
  {
   "backend": "",
   "build": "",
   "chip": "Quadro P5000",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 302.68,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1106",
   "write_tps": 14.86
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Max 10P+4E+30GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 238.76,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/789",
   "write_tps": 20.77
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Max 8P+4E+38GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 225.57,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1291",
   "write_tps": 20.59
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Pro 6P+12E+20GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 216.88,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3180",
   "write_tps": 24.2
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Pro 6P+12E+20GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 218.28,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3226",
   "write_tps": 24.0
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Max 8P+4E+38GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 222.77,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1577",
   "write_tps": 21.96
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Pro 6P+12E+20GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 215.71,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3461",
   "write_tps": 23.73
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Max 12P+4E+40GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 232.57,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1469",
   "write_tps": 17.19
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Max 10P+4E+30GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 215.74,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/805",
   "write_tps": 19.38
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Max 8P+2E+32GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 196.91,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/597",
   "write_tps": 20.17
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Max 8P+2E+32GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 191.39,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/364",
   "write_tps": 19.4
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Pro 5P+10E+16GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 176.78,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3323",
   "write_tps": 22.52
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Pro 10P+4E+20GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 190.44,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/561",
   "write_tps": 17.98
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX A4500 Laptop GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 302.25,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1268",
   "write_tps": 7.94
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 Pro 5P+10E+16GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 169.19,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3531",
   "write_tps": 21.64
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Pro 10P+4E+20GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 186.08,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/777",
   "write_tps": 17.57
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Pro 10P+4E+20GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 182.98,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/369",
   "write_tps": 18.01
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Max 8P+2E+24GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 154.75,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1943",
   "write_tps": 18.19
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Pro 8P+4E+16GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 156.61,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/550",
   "write_tps": 16.71
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Pro 8P+4E+16GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 151.19,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/380",
   "write_tps": 16.1
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 Pro 8P+4E+16GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 151.01,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2277",
   "write_tps": 16.13
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Max 8P+4E+30GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 149.0,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1473",
   "write_tps": 14.49
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Max 8P+2E+24GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 140.19,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/688",
   "write_tps": 15.14
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Pro 8P+4E+19GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 137.33,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/470",
   "write_tps": 14.12
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Pro 6P+6E+18GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 147.29,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/428",
   "write_tps": 11.92
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Pro 6P+6E+18GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 145.35,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/401",
   "write_tps": 11.8
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 Pro 6P+4E+16GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 118.38,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1220",
   "write_tps": 13.34
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Pro 5P+6E+14GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 119.78,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/862",
   "write_tps": 12.1
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4070 Laptop GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 288.99,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1537",
   "write_tps": 3.84
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 Pro 5P+6E+14GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 114.99,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/263",
   "write_tps": 11.61
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 4P+6E+10GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 110.42,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2912",
   "write_tps": 11.54
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Pro 8P+2E+16GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 106.66,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/929",
   "write_tps": 11.91
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Pro 8P+2E+16GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 103.46,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/4",
   "write_tps": 11.6
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Pro 6P+2E+14GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 92.49,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/412",
   "write_tps": 10.83
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Pro 8P+2E+14GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 88.87,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1602",
   "write_tps": 10.4
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 4P+6E+10GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 92.13,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1085",
   "write_tps": 9.08
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 4P+6E+10GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 87.18,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/32",
   "write_tps": 8.99
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M5 4P+6E+10GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 81.11,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3811",
   "write_tps": 8.85
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 Pro 6P+2E+14GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 82.89,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/418",
   "write_tps": 8.93
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 4P+6E+10GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 79.34,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/546",
   "write_tps": 8.59
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX PRO 2000 Blackwell Generation Laptop GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 229.95,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1487",
   "write_tps": 2.14
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 4P+4E+10GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 72.43,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/499",
   "write_tps": 7.69
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 4P+4E+10GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 70.26,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/202",
   "write_tps": 7.35
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon 780M Graphics",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 99.12,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3235",
   "write_tps": 3.05
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M4 4P+6E+8GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 60.89,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1394",
   "write_tps": 7.26
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M3 4P+4E+10GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 65.8,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/562",
   "write_tps": 6.13
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M2 4P+4E+8GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 59.98,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1939",
   "write_tps": 6.96
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 4P+4E+8GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 53.27,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/432",
   "write_tps": 5.43
  },
  {
   "backend": "",
   "build": "",
   "chip": "Apple M1 4P+4E+7GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 40.61,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/223",
   "write_tps": 4.81
  },
  {
   "backend": "",
   "build": "",
   "chip": "AMD Radeon 780M Graphics",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 57.34,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2943",
   "write_tps": 1.94
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3060 Laptop GPU",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 82.9,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/767",
   "write_tps": 1.09
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 4416.36,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/937",
   "write_tps": 64.05
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4090",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 4208.36,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1704",
   "write_tps": 68.84
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4090 D",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 3905.12,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/970",
   "write_tps": 70.46
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX 6000 Ada Generation",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 3552.28,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/673",
   "write_tps": 70.04
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA H100 PCIe",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2973.34,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/69",
   "write_tps": 72.17
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4090",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 3506.19,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/77",
   "write_tps": 49.1
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5090",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 3678.27,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/155",
   "write_tps": 45.54
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5090",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 3789.6,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1125",
   "write_tps": 37.95
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4080 SUPER",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2802.55,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2949",
   "write_tps": 46.15
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3090 Ti",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2202.06,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/127",
   "write_tps": 64.18
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA A100-SXM4-80GB",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2068.63,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/61",
   "write_tps": 66.34
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4080 SUPER",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2680.78,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/102",
   "write_tps": 43.64
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4090",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 3055.72,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1553",
   "write_tps": 31.57
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX 6000 Ada Generation",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2935.63,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/86",
   "write_tps": 28.1
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3090",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1950.01,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1",
   "write_tps": 56.19
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX A6000",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1882.69,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/50",
   "write_tps": 51.97
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4070 Ti",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2176.33,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/138",
   "write_tps": 40.18
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4080",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2339.45,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/15",
   "write_tps": 35.25
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5080",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2415.7,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1119",
   "write_tps": 33.06
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA L40",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2712.93,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/82",
   "write_tps": 24.72
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4070 Ti SUPER",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2146.84,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/136",
   "write_tps": 36.82
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA H100 80GB HBM3",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2608.92,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/87",
   "write_tps": 23.41
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5070 Ti",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2192.57,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/160",
   "write_tps": 36.97
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA A40",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1825.05,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/49",
   "write_tps": 46.17
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3080 Ti",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1792.95,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/120",
   "write_tps": 46.12
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 3080",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1741.04,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/2948",
   "write_tps": 46.62
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA L40S",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2867.59,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/84",
   "write_tps": 19.28
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX A5000",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1532.34,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/26",
   "write_tps": 50.43
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA H100 NVL",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2353.72,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/78",
   "write_tps": 20.48
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA A40-48Q",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1707.41,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3109",
   "write_tps": 39.52
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX PRO 6000 Blackwell Server Edition",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2580.79,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1524",
   "write_tps": 17.94
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 4070",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1691.58,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/147",
   "write_tps": 37.88
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5080",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 2097.1,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/489",
   "write_tps": 25.52
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5070",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1757.7,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/3684",
   "write_tps": 45.69
  },
  {
   "backend": "",
   "build": "",
   "chip": "Tesla V100-SXM2-16GB",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1183.64,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/775",
   "write_tps": 52.97
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA GeForce RTX 5070 Ti",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1794.8,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/1847",
   "write_tps": 25.6
  },
  {
   "backend": "",
   "build": "",
   "chip": "NVIDIA RTX A4500",
   "gpu_cores": null,
   "model": "Qwen2.5 14B Instruct",
   "prompt_tps": 1330.92,
   "quant": "Q4_K_M",
   "retrieved": "2026-10-22",
   "source": "LocalScore",
   "source_url": "https://www.localscore.ai/accelerator/51",
   "write_tps": 43.98
  }
 ]
}