DeepSeek-V4-Flash-Vision-Exp (Max): benchmark results
Provider: DeepSeek. Released 2026-08-21. Access: Open.
Aggregate Rating 1770 ± 14, rank #206 of 2201 rated models, from 131 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Nejumi 4 - jaster (2-shot) - JCoLA (out-of-domain) | 89 | Exact match (%) | 94.9 |
| Nejumi 4 - jaster (0-shot) - CommonsenseMoralJA | 97 | Exact match (%) | 93.5 |
| Nejumi 4 - ARC-AGI-1 | 98 | Accuracy (%) | 93.2 |
| Korean CSAT 2026 (Easy Mode) - Korean | 100 | Points (out of 100) | 92.7 |
| AA GPQA Diamond | 91.31 | Accuracy (%) | 92.2 |
| Korean CSAT 2026 (Easy Mode) - English | 100 | Points (out of 100) | 90.7 |
| Nejumi 4 - BFCL - Relevance Detection | 88.89 | Accuracy (%) | 89.7 |
| AA Long Context Reasoning | 81.33 | Accuracy (%) | 88.9 |
| Nejumi 4 - jaster (0-shot) - NIILC | 57.72 | Character F1 (x100) | 87.7 |
| Nejumi 4 - HLE (Japanese) - Calibration Error | 53 | Calibration error (%) | 87 |
| Tau3 Banking | 41.03 | Success Rate (%) | 86.8 |
| Nejumi 4 - HLE (Japanese) - Accuracy | 38.14 | Accuracy (%) | 86.3 |
Interactive version: theaggregate.ai/model?slug=deepseek-v4-flash-vision-exp-max · How It Works · Data refreshed daily, snapshot 2026-11-12.