stablelm-2-1.6B: benchmark results
Stability AI's Stable LM 2 1.6B, a decoder-only base model pretrained on 2 trillion tokens of multilingual and code data for two epochs, using a LLaMA-like architecture with partial rotary embeddings. Provider: Stability AI. Released 2024-01-19. Access: Open.
Aggregate Rating 1254 ± 27, rank #1536 of 1656 rated models, from 40 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EuroEval Portuguese NLU - SST-2 PT | 75.29 | Matthews correlation coefficient (%) | 46.4 |
| EuroEval Dutch NLU - DBRD | 85.84 | Matthews correlation coefficient (%) | 45.9 |
| EuroEval Norwegian Knowledge - Idioms NO | 2.11 | MCC (x100) | 43.5 |
| EuroEval Portuguese NLU - MultiWikiQA PT | 65.49 | Character F1 (%) | 40.3 |
| Open LLM Leaderboard v1 - GSM8K | 17.44 | Accuracy (%) (5-shot) | 39 |
| Open LLM Leaderboard - MuSR | 38.82 | Score | 34.1 |
| EuroEval Portuguese NLU | 41.85 | NLU Average Score (%) | 32.1 |
| EuroEval Danish Knowledge | 19.62 | Knowledge Average Score (%) | 28.2 |
| EuroEval English Knowledge | 40.83 | Knowledge Average Score (%) | 26.2 |
| EuroEval Portuguese | 28.87 | Average Score (%) | 25.6 |
| EuroEval Danish Knowledge - Danish Citizen Tests | 19.62 | MCC (x100) | 25.4 |
| EuroEval Norwegian Knowledge | 3.25 | Knowledge Average Score (%) | 22 |
Interactive version: theaggregate.ai/model?slug=stablelm-2-1-6b · How It Works · Data refreshed daily, snapshot 2026-10-20.