stablelm-3B-4e1t: benchmark results
Stability AI's StableLM-3B-4E1T, a 3B decoder-only base model pretrained on 1 trillion tokens of English and code data for 4 epochs, using a LLaMA-like architecture with partial rotary embeddings. Provider: Stability AI. Released 2023-09-29. Access: Open.
Aggregate Rating 1237 ± 33, rank #1558 of 1656 rated models, from 26 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EuroEval Portuguese NLU - MultiWikiQA PT | 68.83 | Character F1 (%) | 50.7 |
| EuroEval Portuguese NLU - SST-2 PT | 75.2 | Matthews correlation coefficient (%) | 46.1 |
| EuroEval Portuguese NLU | 44.45 | NLU Average Score (%) | 37.2 |
| EuroEval Portuguese | 30.82 | Average Score (%) | 28.3 |
| Open LLM Leaderboard - MuSR | 37.78 | Score | 28.2 |
| Open LLM Leaderboard v1 - HellaSwag | 75.94 | Normalized accuracy (%) (10-shot) | 27.6 |
| Open LLM Leaderboard v1 - WinoGrande | 71.19 | Accuracy (%) (5-shot) | 26.1 |
| Open Portuguese LLM - BLUEX | 29.9 | Accuracy (%) | 25.8 |
| EuroEval Portuguese NLU - HAREM | 32.09 | Micro F1 excluding MISC (%) | 25.3 |
| Open Portuguese LLM - ENEM | 31.63 | Accuracy (%) | 24.9 |
| Open LLM Leaderboard v1 - MMLU | 45.23 | Accuracy (%) (5-shot) | 23.8 |
| EuroEval Portuguese NLU - ScaLA PT | 1.7 | Matthews correlation coefficient (%) | 23.2 |
Interactive version: theaggregate.ai/model?slug=stablelm-3b-4e1t · How It Works · Data refreshed daily, snapshot 2026-10-20.