| Benchmark | Claude 4 Opus | DeepSeek-V4-Flash |
|---|---|---|
| AA Intelligence | 31.7 | 51.8 |
| AA-LCR | 40 | 74.3 |
| arc_agi_1 | 27 | 84 |
| ARC-AGI-2 | 8.6 | 46 |
| Artificial Analysis Coding Index | 34 | 69.1 |
| GPQA Diamond | 79.6 | 90.8 |
| HLE | 12.3 | 38.6 |
| IFBench | 53.7 | 79.2 |
| LiveCodeBench | 70.4 | 91.6 |
| MMLU-Pro | 86.6 | 86.4 |
| scicode | 40.9 | 49.9 |
| simplebench | 58.8 | 46.3 |
| SWE-bench Verified | 79.4 | 79 |
| Terminal-Bench Hard | 31.1 | 38.6 |
| τ²-Bench Telecom (AA run) | 73.4 | 95.6 |
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.