Benchmarks BENCHMARK DISCUSSION

SWE-bench Verified (OpenAI 477 Tasks)

SWE-bench Verified (OpenAI 477 Tasks) — official/provider-reported evaluation result. Methodology conditions are preserved in the result notes.