{
  "model": "claude-opus-4-6",
  "model_snapshot": "Claude Opus 4.6 (Thinking)",
  "provider": "Anthropic via Antigravity",
  "date_utc": "2026-08-12",
  "timezone": "UTC+01:00",
  "trials_completed": 2,
  "trials_required": 3,
  "fixtures": 12,
  "tools_allowed": [],
  "internet_allowed": false,
  "private_repository_access": false,
  "run_started_utc": "2026-08-12T03:16:56Z",
  "run_ended_utc": "2026-08-12T03:21:22Z",
  "trial_1_status": "Not produced. Trial 1 hit RESOURCE_EXHAUSTED (HTTP/code 429), an API quota limit in the Antigravity API, before output completion. No file was created for trial 1; it is not fabricated or reconstructed here, and this submission publishes only the two trials that actually completed.",
  "raw_files": [
    "raw/claude-opus-4-6-trial-02.json",
    "raw/claude-opus-4-6-trial-03.json"
  ],
  "evidence_status": "finding_not_a_score",
  "why_no_score": "Two reasons, either one sufficient on its own: (1) the protocol requires three independent trials and only two completed, so the trial-count floor is not met; (2) chess-v0.1's answer key is confirmed wrong on 6 of 12 fixtures (see docs/CHESS-BENCHMARK-V0.1.md#wrong-key-the-six-broken-fixtures), so a pass/fail score computed against that key would not measure the model's capability, only its agreement with a mistake. This submission is published as a finding about the benchmark, not as a model result.",
  "raw_files_note": "Copied byte-for-byte from the source run folder; sha256 verified identical before and after copy. Not normalized, repaired, or re-run."
}