SmolLM-1.7B: benchmark results
Hugging Face's SmolLM-1.7B, the largest of three SmolLM sizes (135M, 360M, 1.7B), trained on the Cosmo-Corpus dataset combining synthetic textbooks, Python code and educational web text. Provider: HuggingFace. Released 2024-07-16. Access: Open.
Aggregate Rating 1197 ± 37, rank #1588 of 1661 rated models, from 39 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AI Energy Score (Text Generation) | 4 | Energy Score (1-5) | 47.5 |
| Open PL LLM - CBD (generative, 5-shot) | 15.99 | Macro F1 (%) | 20.9 |
| Open PL LLM - PSC (generative, 5-shot) | 67.12 | Binary F1 (%) | 18.2 |
| Open PL LLM - PSC (multiple choice, 5-shot) | 32.58 | Binary F1 (%) | 17.9 |
| Open Portuguese LLM - OAB Exams | 25.74 | Accuracy (%) | 17.4 |
| Open LLM Leaderboard - IFEval | 23.62 | Score | 17.2 |
| Open Portuguese LLM - FaQuAD NLI | 43.97 | Macro F1 (%) | 17.1 |
| Open Portuguese LLM - BLUEX | 22.81 | Accuracy (%) | 15.4 |
| Open PL LLM - KLEJ NER (generative, 5-shot) | 28.23 | Accuracy (%) | 14.8 |
| Open Portuguese LLM - ASSIN2 RTE | 40.58 | Macro F1 (%) | 12.5 |
| Open PL LLM - RAG | 7.53 | Average RAG Score (%) | 12.2 |
| Open LLM Leaderboard - MATH Level 5 | 1.66 | Score | 12.1 |
Interactive version: theaggregate.ai/model?slug=smollm-1-7b · How It Works · Data refreshed daily, snapshot 2026-10-23.