OpenHermes-2.5-Mistral-7B: benchmark results
Teknium's OpenHermes-2.5-Mistral-7B, a Mistral 7B fine-tune continuing OpenHermes 2 with added code-instruction data, trained on about 1,000,000 mostly GPT-4-generated examples using the ChatML format. Provider: Teknium. Released 2023-10-29. Access: Open.
Aggregate Rating 1399 ± 18, rank #1227 of 1656 rated models, from 69 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open PL LLM - PSC (multiple choice, 5-shot) | 82.54 | Binary F1 (%) | 74.9 |
| Open PL LLM - KLEJ NER (generative, 5-shot) | 52.77 | Accuracy (%) | 72.5 |
| Open PL LLM - PolEmo2-OUT (multiple choice, 5-shot) | 73.28 | Accuracy (%) | 71.3 |
| Open PL LLM - PSC (generative, 5-shot) | 94.46 | Binary F1 (%) | 68.5 |
| Open LLM Leaderboard - IFEval | 55.71 | Score | 67.5 |
| Open PL LLM - DYK (generative, 5-shot) | 61.58 | Binary F1 (%) | 66.4 |
| Open PL LLM - KLEJ NER (multiple choice, 5-shot) | 50.83 | Accuracy (%) | 66.2 |
| Open LLM Leaderboard - MuSR | 42.42 | Score | 64.6 |
| Open PL LLM - DYK (multiple choice, 5-shot) | 57.14 | Binary F1 (%) | 63.8 |
| Open PL LLM - PoQuAD Open Book (generative, 5-shot) | 65.27 | Levenshtein Similarity (%) | 63 |
| Open Portuguese LLM - FaQuAD NLI | 71.2 | Macro F1 (%) | 62.9 |
| Open PL LLM - Multiple Choice | 47.7 | Average Multiple-Choice Score (%) | 62.2 |
Interactive version: theaggregate.ai/model?slug=openhermes-2-5-mistral-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.