| Benchmark | DeepSeek-V4-Pro | o4-mini |
|---|---|---|
| AA Agentic Index | 63.3 | 36.1 |
| AA Intelligence | 53 | 26.1 |
| AA-LCR | 70 | 60 |
| AA-Omniscience | -10.6 | -35.7 |
| Artificial Analysis Coding Index | 59.4 | 25.6 |
| browsecomp | 83.4 | 51.5 |
| critpt | 13 | 0.6 |
| gdpval | 49 | 25.4 |
| GPQA Diamond | 90.5 | 81.4 |
| HLE | 48.2 | 16.5 |
| IFBench | 76.5 | 68.7 |
| LiveCodeBench | 93.5 | 84.5 |
| OmniScience Accuracy | 43 | 24.8 |
| OmniScience Non-Hallucination | 12.2 | 19.5 |
| scicode | 50 | 46.5 |
| simplebench | 50.9 | 38.7 |
| simpleqa_verified | 57.9 | 23.9 |
| strongreject | 98.6 | 96 |
| SWE-bench Verified | 80.6 | 68.1 |
| Terminal-Bench Hard | 46.2 | 15.2 |
| vectara_answer_rate | 97.2 | 99.2 |
| vectara_avg_summary_length | 153.8 | 130.9 |
| vectara_factual_consistency | 91.4 | 81.4 |
| vectara_hallucination_rate ↓ | 8.6 | 18.6 |
| τ²-Bench Telecom (AA run) | 96.2 | 55.6 |
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.