stablelm-2-zephyr-1.6B: benchmark results
Stability AI's StableLM 2 Zephyr 1.6B, an instruction-tuned fine-tune of StableLM 2 1.6B using DPO in a pipeline inspired by Hugging Face's Zephyr 7B, under a non-commercial research license. Provider: Stability AI. Released 2024-01-19. Access: Open.
Aggregate Rating 1299 ± 28, rank #1485 of 1656 rated models, from 23 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| RewardBench Chat | 96.65 | Accuracy (%) | 77.8 |
| Open LLM Leaderboard v1 - GSM8K | 35.33 | Accuracy (%) (5-shot) | 51.1 |
| Open Portuguese LLM - ASSIN2 RTE | 86.51 | Macro F1 (%) | 37.9 |
| Open Portuguese LLM - FaQuAD NLI | 50.99 | Macro F1 (%) | 32.5 |
| Open LLM Leaderboard - IFEval | 32.79 | Score | 30.7 |
| Open Portuguese LLM - OAB Exams | 33.39 | Accuracy (%) | 30.2 |
| Open Portuguese LLM - ENEM | 40.52 | Accuracy (%) | 30 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 45.11 | MC2 (%) (0-shot) | 29.7 |
| Open Portuguese LLM - BLUEX | 32.41 | Accuracy (%) | 28.3 |
| RewardBench | 65.74 | Score (%) | 22.7 |
| Open LLM Leaderboard v1 - MMLU | 42.03 | Accuracy (%) (5-shot) | 21.9 |
| RewardBench Chat Hard | 46.71 | Accuracy (%) | 21.6 |
Interactive version: theaggregate.ai/model?slug=stablelm-2-zephyr-1-6b · How It Works · Data refreshed daily, snapshot 2026-10-20.