{
  "schema": "UNITIQX-LOCAL-MODEL-BENCH-CAPABILITY-v1",
  "generated_utc": "2026-10-11T08:37:00.380506+00:00",
  "execution": "private CLI only; public page does not run inference",
  "endpoint": "local Ollama through Docker bridge",
  "models_found": 9,
  "note": "8 generative/vision tags plus 1 embedding; each model support varies",
  "mode_strategies": [
    "direct",
    "critical",
    "creative",
    "structured",
    "verification"
  ],
  "native_reasoning_mode_control": false,
  "limits": {
    "workers_default": 1,
    "workers_max": 2,
    "models_per_batch_max": 4,
    "tests_per_batch_max": 8,
    "timeout_range_seconds": [
      5,
      120
    ]
  },
  "last_smoke": {
    "tasks": 2,
    "responded": 2,
    "measurements": [
      {
        "model": "qwen3:0.6b",
        "status": "responded",
        "latency_seconds": 14.077,
        "generated_tokens": 170
      },
      {
        "model": "llama3.2:1b",
        "status": "responded",
        "latency_seconds": 8.379,
        "generated_tokens": 92
      }
    ]
  },
  "privacy": "Raw prompts, responses, and run receipts are only saved privately on VPS",
  "warning": "Inference timing is not accuracy; independent grading still needed"
}