| Benchmark | DeepSeek-V3 | o1 |
|---|---|---|
| AA Agentic Index | 1.6 | 31.1 |
| AA Intelligence | 15.4 | 24 |
| AA-LCR | 41.3 | 63.3 |
| AA-Omniscience | -37.6 | -10.6 |
| aider_polyglot | 55.1 | 61.7 |
| AIME 2024 | 59.4 | 79.2 |
| AIME 2025 | 51.3 | 81.7 |
| AIR-Bench 2024 | 40.8 | 80 |
| anthropic_red_team | 97.1 | 98.3 |
| Artificial Analysis Coding Index | 23 | 39.7 |
| bbq | 96.7 | 97.3 |
| Codeforces | 1134 | 2061 |
| Codeforces (Percentile) | 58.7 | 96.6 |
| Codeforces (Rating) | 1134 | 2061 |
| critpt | 0 | 0.3 |
| DROP (3-shot F1) | 91.6 | 90.2 |
| gdpval | 0 | 11.5 |
| GPQA Diamond | 68.4 | 75.7 |
| GSM8K | 96.7 | 97.1 |
| harmbench | 49.7 | 96.3 |
| HLE | 5.2 | 7.7 |
| humaneval | 92.1 | 88.1 |
| IFBench | 41 | 70.3 |
| livebench | 72.4 | 52.3 |
| livecodebench_pass1cot | 40.5 | 63.4 |
| MATH | 91.2 | 96.4 |
| math_500_em | 94 | 96.4 |
| mgsm | 79.8 | 90.8 |
| mmlu | 89.4 | 91.8 |
| mmmlu | 79.4 | 87.7 |
| OmniScience Accuracy | 25.4 | 34.7 |
| OmniScience Non-Hallucination | 23.3 | 30.7 |
| scicode | 35.8 | 35.8 |
| simple_safety_tests | 95.3 | 99 |
| simplebench | 27.2 | 41.7 |
| simpleqa | 27.7 | 47 |
| SWE-bench Verified | 42 | 48.9 |
| Terminal-Bench Hard | 15.2 | 12.9 |
| xstest | 97.1 | 97 |
| τ²-Bench Telecom (AA run) | 47.1 | 62.6 |
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.