| Benchmark | Claude 4 Opus | Claude Opus 4.6 |
|---|---|---|
| AA Intelligence | 31.7 | 44.9 |
| AA-LCR | 40 | 74.3 |
| AIME 2025 | 75.5 | 99.8 |
| arc_agi_1 | 27 | 93 |
| ARC-AGI-2 | 8.6 | 68.8 |
| arena_vision | 1207 | 1300 |
| Artificial Analysis Coding Index | 34 | 48.1 |
| Fortress | 27.6 | 20.5 |
| frontiermath_tier_4 | 4.2 | 22.9 |
| GPQA Diamond | 79.6 | 91.3 |
| HLE | 12.3 | 53.1 |
| IFBench | 53.7 | 62.5 |
| livebench | 74.6 | 76.3 |
| LiveCodeBench | 70.4 | 88.8 |
| LiveCodeBench v6 | 47.4 | 88.8 |
| metr_hcast | 61.7 | 77 |
| metr_re_bench | 20 | 97.4 |
| metr_swaa | 99.8 | 98.5 |
| MMLU-Pro | 86.6 | 89.1 |
| mmmlu | 88.8 | 91.1 |
| multichallenge | 58.6 | 56 |
| MultiNRC | 33.9 | 57.1 |
| scicode | 40.9 | 52 |
| SEAL VISTA | 47 | 46.1 |
| simplebench | 58.8 | 67.6 |
| swe_bench_bash | 67.6 | 75.6 |
| SWE-bench Verified | 79.4 | 80.8 |
| Tau2 retail | 81.8 | 91.9 |
| Terminal-Bench Hard | 31.1 | 65.4 |
| vectara_answer_rate | 91 | 99.8 |
| vectara_avg_summary_length | 123.2 | 137.6 |
| vectara_factual_consistency | 88 | 87.8 |
| vectara_hallucination_rate ↓ | 12 | 12.2 |
| τ²-Bench Telecom (AA run) | 73.4 | 99.3 |
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.