DeepSeek V4 Pro (0813) (Max): benchmark results
Provider: DeepSeek. Released 2026-08-13. Access: Open.
Unified ELO 1703 ± 1, rank #93 of 1761 rated models, from 33 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Vals AI SWE-bench Verified | 96.4 | Resolved (%) | 98.9 |
| OTIS Mock AIME 2024-25 | 98.61 | Accuracy (%) | 95.8 |
| Chess Puzzles (Epoch AI) | 47 | Accuracy (%) | 94.5 |
| Epoch AI - Mystery Game Puzzles | 43 | Score | 94.4 |
| Vals AI Vibe Code Bench | 82.3 | Accuracy (%) | 92.3 |
| Vals AI LiveCodeBench | 87.53 | Accuracy (%) | 91.5 |
| SuperCLUE-Terminal - Overall | 51.52 | Score | 90 |
| Vals AI GPQA | 92.42 | Accuracy (%) | 88.7 |
| LLM2014 Logic 2026-08 | 59.63 | Median Score | 84.1 |
| FrontierMath - Tiers 1-3 (v2) | 64.56 | Accuracy (%, 285 private v2 problems) | 82.3 |
| SimpleQA Verified | 52.91 | Accuracy (%) | 80.5 |
| Epoch AI - Critpt | 18 | Score | 79.9 |
Interactive version: theaggregate.ai/model?slug=deepseek-v4-pro-0813-max · How It Works · Data refreshed daily, snapshot 2026-09-05.