DeepSeek V4 Flash (0731) (Max): benchmark results
Provider: DeepSeek. Released 2026-07-31. Access: Open.
Aggregate Rating 1756 ± 12, rank #232 of 2171 rated models, from 159 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HWE-Bench (OpenHands SDK) | 76.5 | Resolved (%) | 100 |
| Nejumi 4 - jaster (2-shot) - JCoLA (out-of-domain) | 91 | Exact match (%) | 100 |
| Nejumi 4 - jaster (0-shot) - CommonsenseMoralJA | 98 | Exact match (%) | 98.3 |
| Nejumi 4 - ARC-AGI-1 | 98 | Accuracy (%) | 93.4 |
| AA GPQA Diamond | 90.81 | Accuracy (%) | 90.9 |
| IndicBankBench - Contradicting Info | 63.64 | Strict pass^3 (%; 77 cases passed out of the 77 cases of the | 90 |
| Nejumi 4 - ARC-AGI-2 | 62 | Accuracy (%) | 89.6 |
| Nejumi 4 - jaster (2-shot) - JSICK | 84 | Exact match (%) | 89.6 |
| Epoch AI - GPQA Diamond | 91.04 | Accuracy (%) | 89.3 |
| Korean CSAT 2026 (Easy Mode) - Mathematics | 100 | Points (out of 100) | 88.7 |
| LLM2014 Logic 2026-07 | 58.8 | Median Score | 88.4 |
| Nejumi 4 - jaster (0-shot) - NIILC | 57.11 | Character F1 (x100) | 86.8 |
Interactive version: theaggregate.ai/model?slug=deepseek-v4-flash-0731-max · How It Works · Data refreshed daily, snapshot 2026-11-06.