| Benchmark | Claude Sonnet 5 | DeepSeek-V4-Pro |
|---|---|---|
| AA Agentic Index | 49.7 | 63.3 |
| AA Intelligence | 55.3 | 53 |
| AA-LCR | 77 | 70 |
| AA-Omniscience | 16.4 | -10.6 |
| arena_elo | 1463 | 1457 |
| arena_text_factuality | 1453 | 1450 |
| Artificial Analysis Coding Index | 71.5 | 59.4 |
| browsecomp | 84.7 | 83.4 |
| coding_arena_elo | 1540 | 1582 |
| critpt | 16.9 | 13 |
| gdpval | 54.8 | 49 |
| GDPval-AA v2 | 1618 | 1307 |
| GPQA Diamond | 91.1 | 90.5 |
| HLE | 57.4 | 48.2 |
| livebench_agentic_coding | 59.4 | 42.6 |
| livebench_coding | 80.7 | 70 |
| livebench_data_analysis | 71.7 | 74.5 |
| livebench_instruction_following | 63.9 | 62.4 |
| livebench_language | 75 | 78.1 |
| livebench_math | 92.9 | 90.7 |
| livebench_reasoning | 88.7 | 82.7 |
| OmniScience Accuracy | 40 | 43 |
| OmniScience Non-Hallucination | 60.6 | 12.2 |
| scicode | 53.6 | 50 |
| simplebench | 60.6 | 50.9 |
| simpleqa_verified | 25 | 57.9 |
| SWE-bench Multilingual | 78.3 | 76.2 |
| SWE-bench Pro | 63.2 | 55.4 |
| SWE-bench Verified | 85.2 | 80.6 |
| TauBench V3 - Banking | 37.3 | 30.1 |
| Terminal-Bench 2.0 | 80.4 | 67.9 |
| Terminal-Bench 2.1 | 80.5 | 72.1 |
| toolathlon | 54.3 | 51.8 |
| usamo_2026 | 79.5 | 60.7 |
| τ³-Bench | 28.2 | 25.8 |
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.