GPT-6 (Max): benchmark results
Provider: OpenAI. Released 2026-09-03. Access: API.
Unified ELO 1805 ± 1, rank #4 of 1761 rated models, from 68 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA MMMU-Pro | 86.88 | Accuracy (%) | 100 |
| ARC-AGI-2 | 95 | Accuracy (%) | 100 |
| AutomationBench | 41.4 | Pass Rate (%) | 100 |
| Chess Puzzles (Epoch AI) | 72 | Accuracy (%) | 100 |
| Epoch AI - Ebr Bench | 76.19 | Score | 100 |
| Epoch AI - Frontiermath Erdos | 2.94 | Score | 100 |
| Epoch AI - Mystery Game Puzzles | 84 | Score | 100 |
| FrontierMath - Tiers 1-3 (v2) | 93.68 | Accuracy (%, 285 private v2 problems) | 100 |
| LiveBench Plot Unscrambling | 86.29 | Score | 100 |
| LiveBench Table Join | 58.37 | Score | 100 |
| SimpleQA Verified | 75.6 | Accuracy (%) | 100 |
| Vals AI Code Migration | 67.74 | Accuracy (%) | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-6-max · How It Works · Data refreshed daily, snapshot 2026-09-05.