SmolLM2-1.7B: benchmark results
A base/pretrained checkpoint for further adaptation, trained on educational web text, code and curated math data. The project publishes its training and local-inference code. Provider: HuggingFace. Released 2024-10-30. Access: Open.
Aggregate Rating 1242 ± 23, rank #1548 of 1656 rated models, from 156 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LA Leaderboard - MGSM Galician | 6 | Exact Match (%) | 80 |
| LA Leaderboard - caBREU Catalan | 15 | BLEU | 79.4 |
| LA Leaderboard - MGSM Spanish | 3.6 | Exact Match (%) | 77.7 |
| LA Leaderboard - Summarization Galician | 5.92 | BLEU | 77.6 |
| LA Leaderboard - MGSM Catalan | 5.2 | Exact Match (%) | 73.1 |
| LA Leaderboard - HumorQA | 58.82 | Accuracy (%) | 70.6 |
| LA Leaderboard - QNLI Basque | 52.1 | Accuracy (%) | 64 |
| LA Leaderboard - RagQuAS | 76.48 | Semantic Answer Similarity (x100) | 52.9 |
| LA Leaderboard - XQuAD Spanish | 54.03 | F1 (%) | 47.7 |
| LA Leaderboard - EusReading Basque | 27.56 | Accuracy (%) | 47.1 |
| LA Leaderboard - XCOPA Basque | 51 | Accuracy (%) | 46.3 |
| LA Leaderboard - BEC2016 Basque | 37.1 | F1 (%) | 44.9 |
Interactive version: theaggregate.ai/model?slug=smollm2-1-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.