SmolLM3-3B — benchmark results

Hugging Face's fully open Apache-2.0 3B model (July 2025) with dual think/no-think reasoning modes, 128K context, and a published training recipe and data mix. Provider: HuggingFace. Released 2025-07-01. Access: Open.

Unified ELO 1410 ± 14, rank #1172 of 1776 rated models, from 150 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
LA Leaderboard - COPA Spanish84.2Accuracy (%)83.1
AI Energy Score (Text Generation)5Energy Score (1-5)79.5
LA Leaderboard - GalCoLA55.12Accuracy (%)79.4
EuroEval Portuguese NLU - MultiWikiQA PT73.09Reading comprehension Score (%)74.7
EuroEval Portuguese NLU - ScaLA PT21.3Linguistic acceptability Score (%)72.5
EuroEval German NLU - Germanquad56.07Reading comprehension Score (%)70.9
EuroEval Portuguese NLU54.22NLU Average Score (%)70.7
LA Leaderboard55.64Average Score (%)69.9
EuroEval Spanish NLU - MLQA ES62.91Reading comprehension Score (%)69.2
EuroEval Portuguese NLU - SST-2 PT79.85Sentiment classification Score (%)64.5
EuroEval French NLU - FQuAD67.79Reading comprehension Score (%)64.1
EuroEval Spanish NLU - ScaLA ES23.42Linguistic acceptability Score (%)62.3

Interactive version: theaggregate.ai/model?slug=smollm3-3b · How the rankings work · Data refreshed daily, snapshot 2026-07-22.