O3 (Low): benchmark results
O3 evaluated at the low reasoning-effort setting. Provider: OpenAI. Released 2025-04-16. Access: API.
Aggregate Rating 1689 ± 26, rank #394 of 2168 rated models, from 24 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| UGI - Writing | 65.79 | Writing Score | 95.2 |
| TALES | 54.8 | Score (% of max) | 94.7 |
| TALES - Jericho | 14.2 | Score (% of max) | 94.7 |
| TALES - ScienceWorld | 88.3 | Score (% of max) | 94.7 |
| TALES - TextWorld | 99.1 | Score (% of max) | 94.7 |
| UNICBench (Text) | 61.1 | Exact-count hit rate (%): share of parsed numeric answers eq | 94.7 |
| UGI - Natural Intelligence | 65.53 | NatInt Score | 94.6 |
| UGI Leaderboard | 52.09 | UGI Score | 92.6 |
| TALES - TextWorld Express | 89.8 | Score (% of max) | 89.5 |
| TALES - ALFWorld | 70 | Score (% of max) | 86.8 |
| UNICBench (Image) | 18.5 | Exact-count hit rate (%): share of parsed numeric answers eq | 85 |
| LLM Chess (Saplin) | 738.57 | Native anchored-opponent Elo (signed points) | 75.1 |
Interactive version: theaggregate.ai/model?slug=o3-low · How It Works · Data refreshed daily, snapshot 2026-10-20.