O3 Mini: benchmark results
OpenAI's compact o3-series reasoning model (January 2025). Provider: OpenAI. Released 2025-01-31. Access: API.
Aggregate Rating 1669 ± 5, rank #272 of 1656 rated models, from 737 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BBEH | 44.8 | Harmonic Mean (%) | 100 |
| CCR-Bench - Content-Format Hard Satisfaction | 16.6 | Hard satisfaction rate (0-1, times 100): share of instructio | 100 |
| CCR-Bench - Formatting and Structuring | 67.3 | Soft satisfaction rate (0-1, times 100) of the formatting an | 100 |
| COLLIE | 98.7 | Score (%) | 100 |
| Can LLMs Falsify? | 8.9 | Counterexample Rate (%) | 100 |
| CodeARC - Annotated | 59.5 | Success Rate (%; descriptive function names, 1,114 functions | 100 |
| CodeARC - Anonymized | 52.7 | Success Rate (%; anonymized function names, 1,114 functions) | 100 |
| ConflictQA - KG Conflict (Complementary) | 37.67 | Exact match (%) on 430 complementary (answers need both sour | 100 |
| ConflictQA - KG Conflict (Non-Complementary) | 38.03 | Exact match (%) on 802 single-hop (non-complementary) questi | 100 |
| GraphARC - Output Graph Is Connected | 98 | Accuracy (%) of answers to questions about whether the graph | 100 |
| GraphARC - Output Graph Is Tree | 99 | Accuracy (%) of answers to questions about whether the graph | 100 |
| GraphARC - Output Graph Node Count | 83 | Accuracy (%) of answers to questions about the number of nod | 100 |
Interactive version: theaggregate.ai/model?slug=o3-mini · How It Works · Data refreshed daily, snapshot 2026-10-20.