OpenHermes-2-Mistral-7B: benchmark results
Teknium's OpenHermes-2-Mistral-7B, a Mistral 7B fine-tune trained on about 900,000 mostly GPT-4-generated instruction examples, converted to the ChatML format via axolotl. Provider: Teknium. Released 2023-10-12. Access: Open.
Aggregate Rating 1394 ± 19, rank #1247 of 1656 rated models, from 22 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - MuSR | 45.2 | Score | 84.4 |
| Open Korean LLM - Ko-IFEval | 35.58 | Strict accuracy, prompt/instruction mean (%) | 73.4 |
| Open Korean LLM - KorNAT-Harmless | 63.91 | Normalized accuracy (%) | 68.2 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 34.8 | Exact match, flexible extract (%) | 68 |
| Open Korean LLM - KorNAT-Helpful | 48.25 | Normalized accuracy (%) | 66.3 |
| Open Ko-LLM Leaderboard | 37.99 | Average (%) | 64.3 |
| Open Korean LLM Leaderboard | 37.99 | Average Score (%) | 64.3 |
| Open LLM Leaderboard - IFEval | 52.86 | Score | 62.8 |
| Open Portuguese LLM - BLUEX | 54.24 | Accuracy (%) | 59.2 |
| Open Korean LLM - Ko-EQ-Bench | 10.04 | EQ-Bench score | 56.2 |
| Open Portuguese LLM - ENEM | 62.7 | Accuracy (%) | 53 |
| Open Portuguese LLM - ASSIN2 RTE | 89.87 | Macro F1 (%) | 52.8 |
Interactive version: theaggregate.ai/model?slug=openhermes-2-mistral-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.