{
  "model_id": "x2",
  "model_tag": "llama3.3:70b",
  "box_class": "cpu-server",
  "started_utc": "2026-08-25T15:25:32Z",
  "nproc": 64,
  "num_predict": 256,
  "prompt_sha256": "90eedd0c53f9554ae3837674504fcb7090013d9a432a653352183c0f25a7ce5c",
  "nvidia_before": "0, 38764 MiB, 0 %",
  "ollama_version": "0.32.15",
  "runs": [
    {
      "ttft_ms": 377.94,
      "first_token_was_thinking": false,
      "wall_s": 95.1302,
      "eval_count": 256,
      "eval_duration_ns": 94750289000,
      "decode_tok_s": 2.702,
      "prompt_eval_count": 505,
      "prompt_eval_duration_ns": 372361000,
      "prefill_tok_s": 1356.211,
      "load_duration_ns": 1333254,
      "total_duration_ns": 95127925391,
      "streamed_chunks": 256,
      "response_chars": 1030,
      "thinking_chars": 0,
      "done_reason": "length",
      "run": 1
    },
    {
      "ttft_ms": 377.23,
      "first_token_was_thinking": false,
      "wall_s": 95.0124,
      "eval_count": 256,
      "eval_duration_ns": 94633504000,
      "decode_tok_s": 2.705,
      "prompt_eval_count": 505,
      "prompt_eval_duration_ns": 371780000,
      "prefill_tok_s": 1358.33,
      "load_duration_ns": 1108046,
      "total_duration_ns": 95010108836,
      "streamed_chunks": 256,
      "response_chars": 1030,
      "thinking_chars": 0,
      "done_reason": "length",
      "run": 2
    },
    {
      "ttft_ms": 378.68,
      "first_token_was_thinking": false,
      "wall_s": 95.2105,
      "eval_count": 256,
      "eval_duration_ns": 94829616000,
      "decode_tok_s": 2.7,
      "prompt_eval_count": 505,
      "prompt_eval_duration_ns": 372838000,
      "prefill_tok_s": 1354.476,
      "load_duration_ns": 1258385,
      "total_duration_ns": 95207585548,
      "streamed_chunks": 256,
      "response_chars": 1030,
      "thinking_chars": 0,
      "done_reason": "length",
      "run": 3
    }
  ],
  "errors": [],
  "quantization": "Q4_K_M",
  "param_size": "70.6B",
  "family": "llama",
  "context_length": 131072,
  "think_mode": false,
  "think_note": "think:false accepted",
  "warmup_discarded": {
    "ttft_ms": 379.56,
    "first_token_was_thinking": false,
    "wall_s": 95.4532,
    "eval_count": 256,
    "eval_duration_ns": 95072102000,
    "decode_tok_s": 2.693,
    "prompt_eval_count": 505,
    "prompt_eval_duration_ns": 374307000,
    "prefill_tok_s": 1349.16,
    "load_duration_ns": 1221164,
    "total_duration_ns": 95451062268,
    "streamed_chunks": 256,
    "response_chars": 1030,
    "thinking_chars": 0,
    "done_reason": "length"
  },
  "ps_after_load": [
    {
      "name": "llama3.3:70b",
      "model": "llama3.3:70b",
      "size": 44102697614,
      "digest": "a6eb4748fd2990ad2952b2335a95a7f952d1a06119a0aa6a2df6cd052a93a3fa",
      "details": {
        "parent_model": "",
        "format": "gguf",
        "family": "llama",
        "families": [
          "llama"
        ],
        "parameter_size": "70.6B",
        "quantization_level": "Q4_K_M"
      },
      "expires_at": "2026-08-25T15:38:40.2306735Z",
      "size_vram": 0,
      "context_length": 4096
    }
  ],
  "size_vram": 0,
  "size_total": 44102697614,
  "cpu_only_receipt": true,
  "peak_runner_rss_kb": 43154948,
  "peak_runner_rss_gib": 41.16,
  "peak_runner_threads": 103,
  "rss_samples": 517,
  "loadavg_during_runs": {
    "min": 29.38,
    "median": 31.83,
    "max": 32.18,
    "mean": 31.54,
    "samples": 517,
    "nproc": 64
  },
  "contention_suspected": false,
  "contention_note": "loadavg consistent with this bench owning the machine",
  "foreign_gpu_procs_during": "0, 38764 MiB, 0 %",
  "derived": {
    "decode_tok_s": {
      "min": 2.7,
      "median": 2.702,
      "max": 2.705,
      "mean": 2.702,
      "n": 3
    },
    "ttft_ms": {
      "min": 377.23,
      "median": 377.94,
      "max": 378.68,
      "mean": 377.95,
      "n": 3
    },
    "wall_s": {
      "min": 95.0124,
      "median": 95.1302,
      "max": 95.2105,
      "mean": 95.118,
      "n": 3
    },
    "prefill_tok_s": {
      "min": 1354.476,
      "median": 1356.211,
      "max": 1358.33,
      "mean": 1356.339,
      "n": 3
    },
    "eval_count": {
      "min": 256,
      "median": 256,
      "max": 256,
      "mean": 256.0,
      "n": 3
    }
  },
  "cold_load": [
    {
      "ttft_ms": 86973.93,
      "first_token_was_thinking": false,
      "wall_s": 86.9761,
      "eval_count": 1,
      "eval_duration_ns": 1000,
      "decode_tok_s": 1000000.0,
      "prompt_eval_count": 505,
      "prompt_eval_duration_ns": 37569920000,
      "prefill_tok_s": 13.442,
      "load_duration_ns": 49398077956,
      "total_duration_ns": 86973518828,
      "streamed_chunks": 1,
      "response_chars": 3,
      "thinking_chars": 0,
      "done_reason": "length",
      "wall_to_first_token_s": 86.9762,
      "ps_confirmed_empty_before": true,
      "load_duration_s": 49.3981
    }
  ],
  "cold_load_note": "PAGE-CACHE-WARM: no sudo to drop caches, so this is warm-disk reload, not a true cold read from NVMe.",
  "nvidia_after": "0, 38764 MiB, 0 %",
  "ps_final": [
    {
      "name": "llama3.3:70b",
      "model": "llama3.3:70b",
      "size": 44102697614,
      "digest": "a6eb4748fd2990ad2952b2335a95a7f952d1a06119a0aa6a2df6cd052a93a3fa",
      "details": {
        "parent_model": "",
        "format": "gguf",
        "family": "llama",
        "families": [
          "llama"
        ],
        "parameter_size": "70.6B",
        "quantization_level": "Q4_K_M"
      },
      "expires_at": "2026-08-25T15:44:57.790178359Z",
      "size_vram": 0,
      "context_length": 4096
    }
  ],
  "finished_utc": "2026-08-25T15:35:02Z"
}