OLMo 3.1 32B: benchmark results
Ai2's Olmo 3.1 refresh of the fully open Olmo 3 32B line, whose extended-RL recipe underpins the Olmo 3.1 Think and Instruct 32B variants (December 2025). Provider: Allen AI. Released 2025-12-12. Access: Open.
Aggregate Rating 1543 ± 13, rank #632 of 1656 rated models, from 420 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EuroEval Dutch NLU - CoNLL NL | 81.64 | Micro F1 excluding MISC (%) | 100 |
| EuroEval Danish NLU - Angry Tweets | 59.97 | Matthews correlation coefficient (%) | 97.2 |
| EuroEval German NLU - GermEval | 74.37 | Micro F1 excluding MISC (%) | 97.2 |
| EuroEval Catalan NLU - WikiANN CA | 75.06 | Micro F1 excluding MISC (%) | 95.7 |
| Medmarks - Med-HALT Reasoning NOTA | 72.07 | Score (%) | 95.5 |
| EuroEval French NLU - Eltec | 72.07 | Micro F1 excluding MISC (%) | 93.9 |
| EuroEval Finnish NLU - Turku NER FI | 71.69 | Micro F1 excluding MISC (%) | 93.8 |
| EuroEval Portuguese NLU - HAREM | 60.41 | Micro F1 excluding MISC (%) | 93.5 |
| EuroEval Romanian Common Sense Reasoning | 58.01 | Common Sense Reasoning Average Score (%) | 93.1 |
| EuroEval English NLU - CoNLL EN | 84.82 | Micro F1 excluding MISC (%) | 92.9 |
| EuroEval Danish NLU - Dansk | 67.47 | Micro F1 excluding MISC (%) | 92.6 |
| EuroEval Bosnian | 55.41 | Average Score (%) | 92.4 |
Interactive version: theaggregate.ai/model?slug=olmo-3-1-32b · How It Works · Data refreshed daily, snapshot 2026-10-20.