DeepSeekDeepSeek-V4
DeepSeek V4 Flash 0731
DeepSeek · DeepSeek-V4. Released Jul 31, 2026.
DeepSeek V4 Flash 0731 is a model from DeepSeek in the DeepSeek-V4 family, released Jul 31, 2026. evals.report tracks 6 reported DeepSeek V4 Flash 0731 benchmark scores across FrontierCode, ARC-AGI-1, ARC-AGI-2, DeepSWE, Terminal-Bench 2.1, Humanity's Last Exam — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.
Open6 results
Benchmark results 6
Compare this model| Benchmark | Category | Score | Metric | Status | Date | |
|---|---|---|---|---|---|---|
| FrontierCode | Coding | 18.8% | weighted score (Main) | Official | Jul 31, 2026 | Details |
| ARC-AGI-1 | Reasoning | 89% | accuracy | Official | Jul 31, 2026 | Details |
| ARC-AGI-2 | Reasoning | 61.39% | accuracy | Official | Jul 31, 2026 | Details |
| DeepSWE | Coding | 54.4% | % resolved | Verified | Jul 31, 2026 | Details |
| Terminal-Bench 2.1 | Agents | 82.7% | task success | Verified | Jul 31, 2026 | Details |
| Humanity's Last Exam | Reasoning | 37.8% | accuracy | Verified | Jul 31, 2026 | Details |