{
  "bench": "power-cap-96gb",
  "arm": "ollama-q4-mistral-600W",
  "mode": "go",
  "started_utc": "2026-08-25T19:36:55Z",
  "ended_utc": "2026-08-25T19:37:08Z",
  "endpoint": "<serving host>",
  "engine": "ollama",
  "model_requested": "mistral-small3.2:24b",
  "models_served": [
    "mistral-small3.2:24b-instruct-2506-fp16",
    "gemma4:26b-a4b-it-bf16",
    "mistral-small3.2:24b-instruct-2506-q8_0",
    "gemma4:26b-a4b-it-q8_0",
    "llama3.3:70b",
    "nemotron3:33b",
    "granite4.1:30b-q8_0",
    "nomic-embed-text:latest",
    "mistral-small3.2:24b",
    "gemma4:26b"
  ],
  "prompt_file": "p512.txt",
  "prompt_sha256": "90eedd0c53f9554ae3837674504fcb7090013d9a432a653352183c0f25a7ce5c",
  "prompt_tokens_measured": null,
  "max_tokens": 256,
  "ignore_eos": false,
  "thinking": "off",
  "temperature": 0.0,
  "seed": 20260825,
  "repeats": 3,
  "warmup_waves": 1,
  "power_window": {
    "samples": 26,
    "power_w_mean": 470.1,
    "power_w_peak": 561.8,
    "vram_mib_peak": 38772.0,
    "csv": "power.csv"
  },
  "rows": [
    {
      "concurrency": 1,
      "waves": 3,
      "requests_ok": 3,
      "requests_failed": 0,
      "ttft_ms_p50": 20.6,
      "ttft_ms_p95": 20.8,
      "ttft_ms_min": 19.8,
      "ttft_ms_max": 20.8,
      "decode_tok_s_per_stream_p50": 94.81,
      "decode_tok_s_per_stream_mean": 94.8,
      "aggregate_tok_s_mean": 94.44,
      "aggregate_tok_s_max": 94.46,
      "completion_tokens_seen": 768,
      "reasoning_chunks_seen": 0,
      "answer_chars_seen": 3108,
      "distinct_reply_shas": 1
    }
  ],
  "provenance": {
    "harness_sha256": "96a8f38dd1bbb89ce24f89339d81019fd91435e06e7be21b8476f76ba9ce857b",
    "harness_host": "<harness host>",
    "python": "3.14.4",
    "platform": "Linux-7.0.0-30-generic-x86_64-with-glibc2.43"
  },
  "_kit_note": "power.csv and throttle.csv in this directory carry UTC, Z-stamped timestamps; see the kit README for what was converted and why."
}