GPT-5.1 (High): benchmark results
GPT-5.1 evaluated at the high reasoning-effort setting. Provider: OpenAI. Released 2025-11-12. Access: API.
Aggregate Rating 1726 ± 8, rank #291 of 2168 rated models, from 267 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BABE (Biology Arena) | 52.31 | Average score (%), the paper's fixed weighting of its strong | 100 |
| BABE (Biology Arena) - Strong Correlation | 51.79 | Score (%) on the strong-correlation triplets, where each que | 100 |
| DSGym - DSPredict Easy - Above Median | 75 | Above Median Rate (%; private leaderboard) | 100 |
| DSGym - DSPredict Easy - Percentile | 60.4 | Mean Leaderboard Percentile (%; private leaderboard) | 100 |
| DSGym - DSPredict Hard - Above Median | 14.3 | Above Median Rate (%; private leaderboard) | 100 |
| DSGym - MLE-Bench Lite - Above Median | 45.45 | Above Median Rate (%) | 100 |
| ESF-Bench - Conditionals | 88.4 | Tagged slot accuracy (%; slots tagged with a COND (Condition | 100 |
| ESF-Bench - Conversation Understanding | 93.1 | Tagged slot accuracy (%; slots tagged with a CU (Conversatio | 100 |
| ESF-Bench - Reasoning | 64.2 | Tagged slot accuracy (%; slots tagged with a REA (Reasoning) | 100 |
| EnterpriseBench Corecraft (Held-Out Training Split) | 36.86 | Task pass rate (%) on the 150 held-out Corecraft tasks of th | 100 |
| MMPCBench - Diagnostic Precision | 1.79 | Judge-panel score (0-2; detected errors only) | 100 |
| MMPCBench - Strategic Effectiveness | 1.95 | Judge-panel score (0-2; detected errors only) | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-5-1-high · How It Works · Data refreshed daily, snapshot 2026-10-20.