O1 (High): benchmark results
O1 evaluated at the high reasoning-effort setting. Provider: OpenAI. Released 2024-12-05. Access: API.
Aggregate Rating 1646 ± 19, rank #536 of 2168 rated models, from 38 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| PaperBench | 26 | Replication Score (%) | 100 |
| CharXiv-D | 88.9 | Score (%) | 96.7 |
| AbstentionBench - underspecified context - GSM8K - F1 | 97.68 | Abstention F1 (%) | 95.5 |
| AbstentionBench - underspecified context - GSM8K - Recall | 95.47 | Abstention Recall (%) | 95.5 |
| CharXiv-R | 55.1 | Validation chart-visible reasoning accuracy (%) | 92.6 |
| Korean CSAT 2026 (Easy Mode) - Korean | 100 | Points (out of 100) | 92.6 |
| AbstentionBench - underspecified context - MMLU Math - Precision | 100 | Abstention Precision (%) | 89.5 |
| COLLIE | 95.3 | Score (%) | 85.7 |
| AbstentionBench - underspecified context - GSM8K - Precision | 100 | Abstention Precision (%) | 84.1 |
| Korean CSAT 2026 (Easy Mode) - English | 98 | Points (out of 100) | 79.7 |
| AbstentionBench - underspecified context - MMLU Math - F1 | 81.25 | Abstention F1 (%) | 73.7 |
| AbstentionBench - underspecified context - MMLU Math - Recall | 68.42 | Abstention Recall (%) | 73.7 |
Interactive version: theaggregate.ai/model?slug=o1-high · How It Works · Data refreshed daily, snapshot 2026-10-20.