O4 Mini (Low): benchmark results
O4 Mini evaluated at the low reasoning-effort setting. Provider: OpenAI. Released 2025-04-16. Access: API.
Unified ELO 1578 ± 1, rank #523 of 1761 rated models, from 44 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HAL SciCode | 9.23 | Accuracy (%) | 96.7 |
| UGI - Natural Intelligence | 45.54 | NatInt Score | 88.5 |
| UGI - Writing | 46.88 | Writing Score | 86.7 |
| HAL AssistantBench | 28.05 | Accuracy (%) | 85.7 |
| HAL USACO | 53.09 | Accuracy (%) | 81.8 |
| FlagEval VQA - Visual Puzzles | 37.1 | Score | 78.3 |
| HAL GAIA Level 3 | 53.85 | Accuracy (%) | 78.1 |
| HAL SWE-bench Verified Mini | 54 | Score (%) | 76.5 |
| FlagEval VQA - Scene Understanding | 75.8 | Score | 73.9 |
| HAL GAIA Level 1 | 71.7 | Accuracy (%) | 71.9 |
| FlagEval VQA - Long-tail Recognition | 40.2 | Score | 69.6 |
| FormationEval | 94.3 | Accuracy (%) | 62 |
Interactive version: theaggregate.ai/model?slug=o4-mini-low · How It Works · Data refreshed daily, snapshot 2026-09-05.