| Benchmark | DeepSeek-R1 | Gemini 3.1 Pro |
|---|---|---|
| AA Agentic Index | 3.1 | 23 |
| AA Intelligence | 18.6 | 47.7 |
| AA-LCR | 56 | 79 |
| AA-Omniscience | -31.3 | 32.9 |
| arc_agi_1 | 21.2 | 98 |
| ARC-AGI-2 | 1.3 | 77.1 |
| Artificial Analysis Coding Index | 24.6 | 68.8 |
| browsecomp | 8.9 | 85.9 |
| Chinese SimpleQA (C-SimpleQA) | 63.7 | 85.9 |
| Codeforces | 2029 | 3052 |
| critpt | 0.6 | 17.7 |
| Fortress | 74.4 | 29.8 |
| gdpval | 1.5 | 23.3 |
| GPQA Diamond | 81 | 94.3 |
| HLE | 17.7 | 51.4 |
| IFBench | 39 | 77.1 |
| LiveCodeBench | 84.4 | 91.7 |
| mmlu | 92.9 | 92.6 |
| MMLU-Pro | 85 | 91 |
| multichallenge | 45 | 71.4 |
| MultiNRC | 27.6 | 64.7 |
| OmniScience Accuracy | 30.7 | 55.3 |
| OmniScience Non-Hallucination | 10.5 | 50.1 |
| scicode | 35.7 | 59 |
| simplebench | 40.8 | 79.6 |
| simpleqa_verified | 27.4 | 77.3 |
| SWE-bench Multilingual | 30.5 | 76.9 |
| SWE-bench Verified | 57.6 | 80.6 |
| TauBench V3 - Banking | 6.4 | 21.4 |
| Terminal-Bench 2.1 | 19.1 | 74 |
| Terminal-Bench Hard | 6.1 | 68.5 |
| vectara_answer_rate | 97 | 99.4 |
| vectara_avg_summary_length | 93.5 | 107.7 |
| vectara_factual_consistency | 88.7 | 89.6 |
| vectara_hallucination_rate ↓ | 11.3 | 10.4 |
| τ²-Bench Telecom (AA run) | 11.4 | 99.3 |
| τ³-Bench | 6.4 | 67.1 |
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.