GPT-5.1-Codex-Max
Announced on openai.com
Benchmark results
| Benchmark | Score | Source check |
|---|---|---|
| SWE-bench Verifiedxhigh reasoning effort; compaction enabled; all 500 tasks | 77.9% | |
| Terminal-Bench 2.0xhigh reasoning effort; compaction enabled; Codex CLI in the Harbor harness | 58.1% |