| Benchmark | GPT-5.1 | Opus 5 |
|---|---|---|
| AA Agentic Index | 32.2 | 59.2 |
| AA Intelligence | 37.5 | 63.1 |
| AA-LCR | 76.7 | 78.7 |
| AA-Omniscience | 5.4 | 37.1 |
| arc_agi_1 | 72.8 | 97.5 |
| ARC-AGI-2 | 17.6 | 90.4 |
| arena_elo | 1439 | 1487 |
| arena_text_factuality | 1445 | 1481 |
| Artificial Analysis Coding Index | 49.4 | 78 |
| browsecomp | 50.8 | 90.8 |
| coding_arena_elo | 1421 | 1662 |
| critpt | 4.9 | 29.1 |
| gdpval | 25 | 67.2 |
| GPQA Diamond | 88.1 | 93.7 |
| healthbench_length_adjusted | 50.9 | 57.8 |
| HLE | 28.5 | 64.7 |
| HLE (with tools) | 42.7 | 64.7 |
| MCP Atlas | 50.1 | 85.8 |
| MMMU-Pro | 75.5 | 84.7 |
| OmniScience Accuracy | 37.7 | 60.9 |
| OmniScience Non-Hallucination | 48.1 | 40.5 |
| scicode | 43.3 | 55.7 |
| simplebench | 53.2 | 80.6 |
| simpleqa_verified | 48.9 | 56.7 |
| SWE-bench Verified | 76.3 | 96 |
| TauBench V3 - Banking | 15.9 | 44.7 |
| Terminal-Bench 2.1 | 52.4 | 89.1 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.