{
  "schema_version": 1,
  "observed_at": "2026-08-23T02:26:55Z",
  "scope": "Configured upstream used by Halo and AntSeed; not a paid Halo consumer call.",
  "method": {
    "endpoint_shape": "OpenAI chat/completions",
    "structured_request": "Strict json_schema returning status=ok and result=42",
    "plain_fallback": "Plain chat returning the expected arithmetic answer",
    "advertised_min_context_tokens": 128000,
    "completion_token_ceiling": null,
    "max_tokens_is_not_context_window": true,
    "concurrency": 2,
    "not_a_benchmark": true,
    "not_an_sla": true
  },
  "summary": {
    "models_advertised": 8,
    "chat_completions_http_200_observed": 8,
    "strict_structured_output_passed": 7,
    "strict_structured_output_degraded": 1
  },
  "models": {
    "claude-opus-5": {
      "chat_completions": "passed",
      "strict_structured_output": "passed",
      "single_sample_latency_ms": 2628,
      "single_sample_total_tokens": 736
    },
    "claude-sonnet-5": {
      "chat_completions": "passed",
      "strict_structured_output": "passed",
      "single_sample_latency_ms": 2139,
      "single_sample_total_tokens": 804
    },
    "deepseek-v4-flash": {
      "chat_completions": "passed",
      "strict_structured_output": "passed",
      "single_sample_latency_ms": 1338,
      "single_sample_total_tokens": 487
    },
    "deepseek-v4-pro": {
      "chat_completions": "passed",
      "strict_structured_output": "passed",
      "single_sample_latency_ms": 1167,
      "single_sample_total_tokens": 478
    },
    "glm-5.2": {
      "chat_completions": "passed",
      "strict_structured_output": "passed",
      "single_sample_latency_ms": 813,
      "single_sample_total_tokens": 303
    },
    "kimi-k2.7-code": {
      "chat_completions": "passed",
      "strict_structured_output": "passed",
      "single_sample_latency_ms": 2148,
      "single_sample_total_tokens": 198
    },
    "minimax-m3": {
      "chat_completions": "passed",
      "strict_structured_output": "passed",
      "single_sample_latency_ms": 2173,
      "single_sample_total_tokens": 586
    },
    "qwen3.5-397b-a17b": {
      "chat_completions": "passed_without_explicit_completion_ceiling",
      "json_object": "passed_with_max_tokens_65536",
      "forced_tool": "passed_without_explicit_completion_ceiling",
      "strict_structured_output": "degraded_http_503",
      "strict_attempts_without_artificial_low_ceiling": 2,
      "plain_single_sample_latency_ms": 6088,
      "plain_single_sample_total_tokens": 237
    }
  },
  "buyer_guidance": {
    "structured_output_preferred_models": [
      "claude-opus-5",
      "claude-sonnet-5",
      "deepseek-v4-flash",
      "deepseek-v4-pro",
      "glm-5.2",
      "kimi-k2.7-code",
      "minimax-m3"
    ],
    "qwen3.5-397b-a17b": "The model has at least a 128000-token context window. Plain chat and forced-tool calls passed without an artificial completion ceiling; json_object passed with max_tokens=65536. Use another model for strict json_schema until the staged shim fix is deployed and reverified."
  },
  "evidence_runs": [
    "RUN-20260823T022608Z-c6c6ec",
    "RUN-20260823T022538Z-fe9c69",
    "RUN-20260823T022550Z-8ee1f5",
    "RUN-20260823T022555Z-d9d5ae",
    "RUN-20260823T022458Z-221f57",
    "RUN-20260823T022628Z-ec87ca",
    "RUN-20260823T022259Z-4fae3b"
  ],
  "superseded_diagnostics": [
    {
      "runs": [
        "RUN-20260823T014314Z-a8486e",
        "RUN-20260823T014352Z-c21d96",
        "RUN-20260823T014425Z-b47568"
      ],
      "reason": "The 32/128/512 max_tokens probes measured artificial completion ceilings, not the model's at-least-128000-token context window."
    }
  ]
}