| Benchmark | DeepSeek-V4-Flash | Gemini 1.5 Pro |
|---|---|---|
| AA Intelligence | 51.8 | 10 |
| ARC-AGI-2 | 46 | 0.8 |
| Artificial Analysis Coding Index | 69.1 | 23.6 |
| Chinese SimpleQA (C-SimpleQA) | 78.9 | 59.4 |
| GPQA Diamond | 90.8 | 59.1 |
| HLE | 38.6 | 4.9 |
| LiveCodeBench | 91.6 | 30.5 |
| MMLU-Pro | 86.4 | 75.8 |
| scicode | 49.9 | 29.5 |
| simplebench | 46.3 | 27.1 |
| SWE-bench Verified | 79 | 34.2 |
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.