{
  "recorded_on": "2026-10-18",
  "status": "measured-lab-only",
  "production_promoted": false,
  "hardware": {
    "gpu": "RTX 5070 Ti 16 GB",
    "ram": "32 GB",
    "cpu": "Intel Core i9-13900K"
  },
  "runtime": "Ollama 0.40.6",
  "runtime_sha256": "71d2002d3ae572d5fa8c1d368ce9d1b9a39c7187cf169e8b330c55410bf6c58e",
  "profiles": 72,
  "repeats": 3,
  "models_sha256": {
    "gemma4:12b": "4eb23ef187e2c5462566d6a1d3bbbc2f1346d0b4327cbb66d58fffbcc9b2b05c",
    "lfm2.5:8b": "9cf756159fc2f3b9128c6a3f544ec90c5e9b8afdbb4179a57b8aea9de589cfb2",
    "qwen3.5:4b": "2a654d98e6fba55d452b7043684e9b57a947e393bbffa62485a7aac05ee4eefd",
    "qwen3.5:9b": "6488c96fa5faab64bb65cbd30d4289e20e6130ef535a93ef9a49f42eda893ea7"
  },
  "source_hashes": {
    "cache_comparison": "c1b55f64373fdefc041ecf5694b86873f0eb32cdf6d385888277b64bfdce86bb",
    "context_plan": "b2f716f9490729126c5f1f8e70ce909b91809fce55108bc41f8987e550e32d9f",
    "cache_plan": "28f59b5b670474a43a365de94c6150fa86a1d7f01f844bd2549c21e5170dc83d",
    "lab_model_selection": "514dd2d3fdbe3f2aa44dea072622b074e0fdfb01b11a8f74f1ba673d57dc2384"
  },
  "method": "Four models, two allocated context capacities, three cache types, three repeats. Per cell: minimum passed tasks, median recorded e2e_tok_s and median peak VRAM bytes. Each candidate is compared with its own model/context f16 reference.",
  "limits": [
    "Short prompts; allocated context is not long-input quality.",
    "e2e_tok_s includes prefill and generation, subtracts model load and excludes server startup and cleanup; not total wall-clock throughput.",
    "Cold means empty model residency, not cleared OS or compiler caches.",
    "All eight model/context reference cells retained f16. LFM scored zero and remains excluded from the selected lab profile.",
    "No production profile was promoted. This is not a model leaderboard or whole-machine speedup."
  ],
  "rows": [
    {"model": "gemma4:12b", "context": 16384, "cache": "f16", "min_passed": 16, "median_e2e_tok_s": 72.1, "median_peak_vram_bytes": 8387611851},
    {"model": "gemma4:12b", "context": 16384, "cache": "q8_0", "min_passed": 16, "median_e2e_tok_s": 68.7, "median_peak_vram_bytes": 8074381229},
    {"model": "gemma4:12b", "context": 16384, "cache": "q4_0", "min_passed": 14, "median_e2e_tok_s": 69.7, "median_peak_vram_bytes": 7906609069},
    {"model": "gemma4:12b", "context": 65536, "cache": "f16", "min_passed": 16, "median_e2e_tok_s": 72.6, "median_peak_vram_bytes": 8488275147},
    {"model": "gemma4:12b", "context": 65536, "cache": "q8_0", "min_passed": 16, "median_e2e_tok_s": 68.6, "median_peak_vram_bytes": 8187627437},
    {"model": "gemma4:12b", "context": 65536, "cache": "q4_0", "min_passed": 14, "median_e2e_tok_s": 69.8, "median_peak_vram_bytes": 8019855277},
    {"model": "lfm2.5:8b", "context": 16384, "cache": "f16", "min_passed": 0, "median_e2e_tok_s": 339.1, "median_peak_vram_bytes": 5504174652},
    {"model": "lfm2.5:8b", "context": 16384, "cache": "q8_0", "min_passed": 0, "median_e2e_tok_s": 328.2, "median_peak_vram_bytes": 5409834269},
    {"model": "lfm2.5:8b", "context": 16384, "cache": "q4_0", "min_passed": 0, "median_e2e_tok_s": 326.6, "median_peak_vram_bytes": 5359502621},
    {"model": "lfm2.5:8b", "context": 65536, "cache": "f16", "min_passed": 0, "median_e2e_tok_s": 339.5, "median_peak_vram_bytes": 6464691239},
    {"model": "lfm2.5:8b", "context": 65536, "cache": "q8_0", "min_passed": 0, "median_e2e_tok_s": 327.5, "median_peak_vram_bytes": 6087245822},
    {"model": "lfm2.5:8b", "context": 65536, "cache": "q4_0", "min_passed": 0, "median_e2e_tok_s": 328.3, "median_peak_vram_bytes": 5885919230},
    {"model": "qwen3.5:4b", "context": 16384, "cache": "f16", "min_passed": 13, "median_e2e_tok_s": 115.8, "median_peak_vram_bytes": 3648153189},
    {"model": "qwen3.5:4b", "context": 16384, "cache": "q8_0", "min_passed": 10, "median_e2e_tok_s": 109.2, "median_peak_vram_bytes": 3434516314},
    {"model": "qwen3.5:4b", "context": 16384, "cache": "q4_0", "min_passed": 13, "median_e2e_tok_s": 113.9, "median_peak_vram_bytes": 3300298586},
    {"model": "qwen3.5:4b", "context": 65536, "cache": "f16", "min_passed": 13, "median_e2e_tok_s": 115.5, "median_peak_vram_bytes": 5648878140},
    {"model": "qwen3.5:4b", "context": 65536, "cache": "q8_0", "min_passed": 10, "median_e2e_tok_s": 109.1, "median_peak_vram_bytes": 4852233009},
    {"model": "qwen3.5:4b", "context": 65536, "cache": "q4_0", "min_passed": 13, "median_e2e_tok_s": 113.1, "median_peak_vram_bytes": 4315362097},
    {"model": "qwen3.5:9b", "context": 16384, "cache": "f16", "min_passed": 15, "median_e2e_tok_s": 89.6, "median_peak_vram_bytes": 6047945194},
    {"model": "qwen3.5:9b", "context": 16384, "cache": "q8_0", "min_passed": 15, "median_e2e_tok_s": 87.0, "median_peak_vram_bytes": 5796559583},
    {"model": "qwen3.5:9b", "context": 16384, "cache": "q4_0", "min_passed": 14, "median_e2e_tok_s": 88.7, "median_peak_vram_bytes": 5662341855},
    {"model": "qwen3.5:9b", "context": 65536, "cache": "f16", "min_passed": 15, "median_e2e_tok_s": 88.9, "median_peak_vram_bytes": 8111584705},
    {"model": "qwen3.5:9b", "context": 65536, "cache": "q8_0", "min_passed": 15, "median_e2e_tok_s": 87.3, "median_peak_vram_bytes": 7205887670},
    {"model": "qwen3.5:9b", "context": 65536, "cache": "q4_0", "min_passed": 14, "median_e2e_tok_s": 88.3, "median_peak_vram_bytes": 6669016758}
  ]
}