| Benchmark | Claude Opus 4.6 | DeepSeek-R1 |
|---|---|---|
| AA Agentic Index | 67.6 | 3.1 |
| AA Intelligence | 44.9 | 18.6 |
| AA-LCR | 74.3 | 56 |
| AA-Omniscience | 13.7 | -31.3 |
| AIME 2025 | 99.8 | 87.5 |
| arc_agi_1 | 93 | 21.2 |
| ARC-AGI-2 | 68.8 | 1.3 |
| Artificial Analysis Coding Index | 48.1 | 24.6 |
| browsecomp | 86.8 | 8.9 |
| Chinese SimpleQA (C-SimpleQA) | 76.4 | 63.7 |
| critpt | 12.6 | 0.6 |
| Fortress | 20.5 | 74.4 |
| gdpval | 55.9 | 1.5 |
| GPQA Diamond | 91.3 | 81 |
| HLE | 53.1 | 17.7 |
| IFBench | 62.5 | 39 |
| LiveCodeBench | 88.8 | 84.4 |
| MMLU-Pro | 89.1 | 85 |
| multichallenge | 56 | 45 |
| MultiNRC | 57.1 | 27.6 |
| OmniScience Accuracy | 47 | 30.7 |
| OmniScience Non-Hallucination | 37.2 | 10.5 |
| scicode | 52 | 35.7 |
| simplebench | 67.6 | 40.8 |
| simpleqa_verified | 46.5 | 27.4 |
| SWE-bench Multilingual | 77.8 | 30.5 |
| SWE-bench Verified | 80.8 | 57.6 |
| Terminal-Bench Hard | 65.4 | 6.1 |
| vectara_answer_rate | 99.8 | 97 |
| vectara_avg_summary_length | 137.6 | 93.5 |
| vectara_factual_consistency | 87.8 | 88.7 |
| vectara_hallucination_rate ↓ | 12.2 | 11.3 |
| τ²-Bench Telecom (AA run) | 99.3 | 11.4 |
| τ³-Bench | 72.4 | 6.4 |
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.