stablelm-2-12B: benchmark results
Stability AI's StableLM 2 12B, a 12.1B decoder-only base model pretrained on 2 trillion tokens of multilingual and code data over two epochs. Provider: Stability AI. Released 2024-04-08. Access: Open.
Aggregate Rating 1404 ± 29, rank #1194 of 1635 rated models, from 17 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - MuSR | 44.79 | Score | 82.4 |
| Open LLM Leaderboard v1 - GSM8K | 56.03 | Accuracy (%) (5-shot) | 67.6 |
| Open Portuguese LLM - OAB Exams | 47.29 | Accuracy (%) | 65.7 |
| Open LLM Leaderboard v1 - HellaSwag | 84.33 | Normalized accuracy (%) (10-shot) | 65.4 |
| Open Portuguese LLM - FaQuAD NLI | 71.14 | Macro F1 (%) | 62.7 |
| Open LLM Leaderboard v1 - WinoGrande | 77.9 | Accuracy (%) (5-shot) | 59.3 |
| Open Portuguese LLM - BLUEX | 54.1 | Accuracy (%) | 58.6 |
| Open Portuguese LLM - ENEM | 64.31 | Accuracy (%) | 58.2 |
| Open LLM Leaderboard v1 - MMLU | 62.04 | Accuracy (%) (5-shot) | 53.8 |
| Open Portuguese LLM - ASSIN2 RTE | 89.86 | Macro F1 (%) | 52.6 |
| Open LLM Leaderboard v1 - ARC Challenge | 58.45 | Normalized accuracy (%) (25-shot) | 39.8 |
| Open LLM Leaderboard - MMLU-Pro | 30.72 | Score | 39 |
Interactive version: theaggregate.ai/model?slug=stablelm-2-12b · How It Works · Data refreshed daily, snapshot 2026-11-05.