| Benchmark | Claude Opus 4.6 | GPT-4o |
|---|---|---|
| AA Agentic Index | 67.6 | 9.7 |
| AA Intelligence | 44.9 | 11.2 |
| AA-LCR | 74.3 | 55 |
| AA-Omniscience | 13.7 | -10.5 |
| AIME 2025 | 99.8 | 11.7 |
| arc_agi_1 | 93 | 4.5 |
| ARC-AGI-2 | 68.8 | 0 |
| arena_vision | 1300 | 1162 |
| Artificial Analysis Coding Index | 48.1 | 24.2 |
| Chinese SimpleQA (C-SimpleQA) | 76.4 | 64.6 |
| critpt | 12.6 | 0 |
| ERQA | 40.8 | 35.2 |
| Fortress | 20.5 | 47.2 |
| gdpval | 55.9 | 0 |
| GPQA Diamond | 91.3 | 70.1 |
| GSM8K | 0.6 | 95.6 |
| HLE | 53.1 | 5.3 |
| IFBench | 62.5 | 36 |
| LiveCodeBench | 88.8 | 38.3 |
| LiveCodeBench v6 | 88.8 | 30.9 |
| mathvision | 71.2 | 30.4 |
| metr_hcast | 77 | 25 |
| metr_re_bench | 97.4 | 0 |
| metr_swaa | 98.5 | 99.2 |
| MMLU-Pro | 89.1 | 74.7 |
| mmmlu | 91.1 | 81.4 |
| MMMU-Pro | 77.3 | 59.9 |
| multichallenge | 56 | 40.3 |
| MultiNRC | 57.1 | 12.4 |
| OmniScience Accuracy | 47 | 23.7 |
| OmniScience Non-Hallucination | 37.2 | 62.1 |
| OSWorld-Verified | 72.7 | 5 |
| RealWorldQA | 73.9 | 75.4 |
| scicode | 52 | 33.4 |
| SEAL VISTA | 46.1 | 34.9 |
| simplebench | 67.6 | 17.8 |
| swe_bench_bash | 75.6 | 21.6 |
| SWE-bench Verified | 80.8 | 38.8 |
| Tau2 retail | 91.9 | 63.4 |
| Terminal-Bench Hard | 65.4 | 8.3 |
| vectara_answer_rate | 99.8 | 93.8 |
| vectara_avg_summary_length | 137.6 | 86.6 |
| vectara_factual_consistency | 87.8 | 90.4 |
| vectara_hallucination_rate ↓ | 12.2 | 9.6 |
| τ²-Bench Telecom (AA run) | 99.3 | 28.9 |
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.