GPT-5 Mini (Medium): benchmark results
GPT-5 Mini evaluated at the medium reasoning-effort setting. Provider: OpenAI. Released 2025-08-07. Access: API.
Aggregate Rating 1633 ± 10, rank #589 of 2168 rated models, from 150 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CAME-Bench - Medium | 56.6 | Answer-set macro F1 (%; 168 questions on six trajectories av | 100 |
| CAME-Bench - Small | 80.4 | Answer-set macro F1 (%; 144 questions on six trajectories av | 100 |
| Elimination Game (Lechmazur) | 5.73 | TrueSkill μ | 96.6 |
| SLR-Bench - Easy | 98 | Accuracy (%) | 95.5 |
| SLR-Bench - Hard | 23 | Accuracy (%) | 91.8 |
| VisReason (Vision-Centric) - Cue Insight | 55.9 | Accuracy (%) on the Cue Insight category (conceptual reasoni | 90.5 |
| VisReason (Vision-Centric) - Pattern Counting | 32.6 | Accuracy (%) on the Pattern Counting category (perceptual re | 90.5 |
| SLR-Bench - Medium | 50 | Accuracy (%) | 89.1 |
| AA IFBench | 71.16 | Accuracy (%) | 86.6 |
| SLR-Bench - Basic | 100 | Accuracy (%) | 86.4 |
| VisReason (Vision-Centric) - Deductive Reasoning | 44.4 | Accuracy (%) on the Deductive Reasoning category (conceptual | 85.7 |
| VisReason (Vision-Centric) - Geolocation | 56.5 | Accuracy (%) on the Geolocation category (conceptual reasoni | 85.7 |
Interactive version: theaggregate.ai/model?slug=gpt-5-mini-medium · How It Works · Data refreshed daily, snapshot 2026-10-20.