OLMo 3 32B: benchmark results
Ai2's fully open 32B flagship of the Olmo 3 family, released with its complete data-to-checkpoint 'model flow' on the Dolma 3 corpus (November 2025). Provider: Allen AI. Released 2025-11-20. Access: Open.
Aggregate Rating 1503 ± 15, rank #772 of 1656 rated models, from 159 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA LiveCodeBench | 67.2 | Pass@1 (%) | 75.1 |
| Swallow - Post-trained Japanese - JHumanEval | 92.1 | Pass@1 (%) | 73.1 |
| Swallow - Post-trained Japanese - PolyMath High and Top | 45.2 | Accuracy (%) | 71.6 |
| Swallow - Post-trained English - AIME | 75 | Accuracy (%) | 70.9 |
| Swallow - Post-trained English - MATH-500 | 96.8 | Accuracy (%) | 69.4 |
| AA AIME 2025 | 73.67 | Accuracy (%) | 68.8 |
| Swallow - Post-trained English - LiveCodeBench | 56.9 | Pass@1 (%) | 67.2 |
| Medmarks - SCTpublic | 67.64 | Score (%) | 67 |
| Medmarks - Med-HALT Reasoning NOTA | 63.53 | Score (%) | 65.9 |
| AA IFBench | 49.12 | Accuracy (%) | 58.4 |
| Swallow - Post-trained English - MMLU-Pro | 76 | Accuracy (%) | 58.2 |
| Medmarks - Med-HALT Reasoning FCT | 71.67 | Score (%) | 55.7 |
Interactive version: theaggregate.ai/model?slug=olmo-3-32b · How It Works · Data refreshed daily, snapshot 2026-10-20.