{
  "schema": 1,
  "id": "apple-m5-24gb-2026-10-24",
  "measured_at": "2026-10-24T06:44:03.252912+00:00",
  "hardware": {
    "chip": "Apple M5",
    "memory_gib": 24,
    "os": "macOS 27.0.1",
    "arch": "arm64"
  },
  "method": {
    "kind": "llama-server-completion",
    "prompt_sha256": "d1073ca91137cbd8b8c777e44e3634b6d02a3f492eaf4805ec47f5f68c9654fb",
    "server_sha256": "4bc6b545e77fc15471493503fe2d63a3e7dfee0a46e6735b2244c8f4eb934bf5",
    "llama_cpp_revision": "5aeac37a35b44fde4bdffaec9159f2fcb253b622",
    "app_commit": "5cadd45b7a78e2c6abec380dcc2f6a89d31d3b66",
    "app_dirty": true,
    "generation_tokens_requested": 128,
    "repetitions": 3,
    "cache": "prompt cache disabled; model warm after one unmeasured 16-token request",
    "rss_sampling_interval_ms": 100,
    "synthetic_input": true
  },
  "runs": [
    {
      "backend": "cpu",
      "model_id": "qwen3-1.7b",
      "model_sha256": "b139949c5bd74937ad8ed8c8cf3d9ffb1e99c866c823204dc42c0d91fa181897",
      "quantization": "Q4_K_M",
      "context_size": 2048,
      "threads": 2,
      "batch_size": 128,
      "ubatch_size": 128,
      "repeat": 1,
      "prompt_tokens": 155,
      "generated_tokens": 128,
      "prompt_tps": 69.94376521276894,
      "generation_tps": 14.33738855643834,
      "ttft_ms": 2229.2705830004707,
      "elapsed_ms": 11087.756208002247,
      "peak_rss_mib": 2426.140625,
      "model_startup_ms": 1155.787625000812
    },
    {
      "backend": "cpu",
      "model_id": "qwen3-1.7b",
      "model_sha256": "b139949c5bd74937ad8ed8c8cf3d9ffb1e99c866c823204dc42c0d91fa181897",
      "quantization": "Q4_K_M",
      "context_size": 2048,
      "threads": 2,
      "batch_size": 128,
      "ubatch_size": 128,
      "repeat": 2,
      "prompt_tokens": 155,
      "generated_tokens": 128,
      "prompt_tps": 70.28299106272418,
      "generation_tps": 13.149203075464003,
      "ttft_ms": 2212.9676250006014,
      "elapsed_ms": 11871.683083001699,
      "peak_rss_mib": 2425.9375,
      "model_startup_ms": 1155.787625000812
    },
    {
      "backend": "cpu",
      "model_id": "qwen3-1.7b",
      "model_sha256": "b139949c5bd74937ad8ed8c8cf3d9ffb1e99c866c823204dc42c0d91fa181897",
      "quantization": "Q4_K_M",
      "context_size": 2048,
      "threads": 2,
      "batch_size": 128,
      "ubatch_size": 128,
      "repeat": 3,
      "prompt_tokens": 155,
      "generated_tokens": 128,
      "prompt_tps": 74.8387707321501,
      "generation_tps": 13.522211296796035,
      "ttft_ms": 2075.5097910005134,
      "elapsed_ms": 11468.28537500187,
      "peak_rss_mib": 2424.34375,
      "model_startup_ms": 1155.787625000812
    },
    {
      "backend": "metal",
      "model_id": "qwen3-1.7b",
      "model_sha256": "b139949c5bd74937ad8ed8c8cf3d9ffb1e99c866c823204dc42c0d91fa181897",
      "quantization": "Q4_K_M",
      "context_size": 2048,
      "threads": 2,
      "batch_size": 128,
      "ubatch_size": 128,
      "repeat": 1,
      "prompt_tokens": 155,
      "generated_tokens": 128,
      "prompt_tps": 1492.1924639467047,
      "generation_tps": 78.32220274103041,
      "ttft_ms": 108.68612500053132,
      "elapsed_ms": 1730.8535000011034,
      "peak_rss_mib": 1431.234375,
      "model_startup_ms": 421.6401660014526
    },
    {
      "backend": "metal",
      "model_id": "qwen3-1.7b",
      "model_sha256": "b139949c5bd74937ad8ed8c8cf3d9ffb1e99c866c823204dc42c0d91fa181897",
      "quantization": "Q4_K_M",
      "context_size": 2048,
      "threads": 2,
      "batch_size": 128,
      "ubatch_size": 128,
      "repeat": 2,
      "prompt_tokens": 155,
      "generated_tokens": 128,
      "prompt_tps": 1612.668289739268,
      "generation_tps": 79.66917844139485,
      "ttft_ms": 101.90095800135168,
      "elapsed_ms": 1696.255208000366,
      "peak_rss_mib": 1431.328125,
      "model_startup_ms": 421.6401660014526
    },
    {
      "backend": "metal",
      "model_id": "qwen3-1.7b",
      "model_sha256": "b139949c5bd74937ad8ed8c8cf3d9ffb1e99c866c823204dc42c0d91fa181897",
      "quantization": "Q4_K_M",
      "context_size": 2048,
      "threads": 2,
      "batch_size": 128,
      "ubatch_size": 128,
      "repeat": 3,
      "prompt_tokens": 155,
      "generated_tokens": 128,
      "prompt_tps": 1631.991239892183,
      "generation_tps": 75.12925781758379,
      "ttft_ms": 99.5325830008369,
      "elapsed_ms": 1790.172207998694,
      "peak_rss_mib": 1431.234375,
      "model_startup_ms": 421.6401660014526
    }
  ],
  "summary": [
    {
      "backend": "cpu",
      "model_id": "qwen3-1.7b",
      "median_generation_tps": 13.522211296796035,
      "median_ttft_ms": 2212.9676250006014,
      "max_peak_rss_mib": 2426.140625
    },
    {
      "backend": "metal",
      "model_id": "qwen3-1.7b",
      "median_generation_tps": 78.32220274103041,
      "median_ttft_ms": 101.90095800135168,
      "max_peak_rss_mib": 1431.328125
    }
  ],
  "limitations": [
    "Single machine; not a prediction for other M-series chips.",
    "Generation speed is not transcription accuracy or meeting-summary quality.",
    "RSS is a sampled process-resident-memory peak, not total unified/GPU allocation.",
    "Power draw and thermal throttling were not measured.",
    "128 output tokens are forced for comparability; this is a short synthetic benchmark."
  ],
  "backend_evidence": [
    {
      "backend": "cpu",
      "gpu_layers": 0,
      "total_layers": 29,
      "log_sha256": "b3b8ce62e4484f0a020957b246e5489c3ee8f0d919090639313124d23eec5023"
    },
    {
      "backend": "metal",
      "gpu_layers": 29,
      "total_layers": 29,
      "log_sha256": "f9d78a409c7d3298f0ab3c0849225328867fe306ee52e6cfd877494b5019df2d"
    }
  ]
}