o3 Mini is a compact reasoning-focused model that punches above its weight on logic, math, and code problems by spending extra compute on chain-of-thought before answering. It trades breadth for depth — it handles structured problem-solving confidently but lacks vision input and the broad general knowledge surface of larger models. Think of it as a focused analyst who works through problems methodically rather than a generalist who knows a little about everything.
| Benchmark | Score | Type | Recorded |
|---|---|---|---|
| TAU2 | 32.4 | accuracy | 5d ago |
| SWE-Bench | 49.3 | accuracy | 5d ago |
| AIME 2024 | 87.3 | accuracy | 5d ago |
| Aider Polyglot | 60.4 | accuracy | 5d ago |
| AIME 2025 | 86.5 | accuracy | 5d ago |
| SciCode | 10.8 | main_problem_pass@1 | 5d ago |