OLMo 3 7B Instruct: benchmark results

Allen AI OLMo 3 7B instruction-tuned checkpoint. Provider: Allen AI. Released 2025-11-20. Access: Open.

Unified ELO 1410 ± 1, rank #1095 of 1392 rated models, from 367 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
CringeBench0.69Cringe Score (0-10, lower is better)87.9
EuroEval Portuguese NLU - ScaLA PT27.42Linguistic acceptability Score (%)82.5
MERA Code - CodeCorrectness81.56Accuracy (%)80
EuroEval English NLU - ScaLA EN52.13Linguistic acceptability Score (%)78.4
EuroEval Italian NLU - ScaLA IT29.13Linguistic acceptability Score (%)76
EuroEval Polish Summarization - PSC22.96Score (%)70.7
EuroEval Dutch Summarization - Wiki Lingua NL33.57Score (%)70
MERA - LCS16Accuracy (%)64.6
EuroEval Catalan Summarization - Dacsa CA34.5Score (%)63.6
EuroEval French NLU - ScaLA FR37.65Linguistic acceptability Score (%)61.7
EuroEval Spanish NLU - ScaLA ES21.12Linguistic acceptability Score (%)59.9
Medmarks - MedHallu Hard45.94Score (%)58.6

Interactive version: theaggregate.ai/model?slug=olmo-3-7b-instruct · How It Works · Data refreshed daily, snapshot 2026-09-05.