Nous-Hermes-2-Mistral-7B-DPO: benchmark results
Nous Research's Nous-Hermes-2-Mistral-7B-DPO, a DPO-aligned version of OpenHermes-2.5-Mistral-7B, itself trained on 1,000,000 mostly GPT-4-quality instructions, using the ChatML prompt format. Provider: Nous Research. Released 2024-01-15. Access: Open.
Aggregate Rating 1416 ± 23, rank #1164 of 1656 rated models, from 38 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - IFEval | 57.63 | Score | 69.5 |
| Open Medical LLM - PubMedQA | 74.8 | Accuracy (%) | 65.9 |
| Open Portuguese LLM - FaQuAD NLI | 71.38 | Macro F1 (%) | 63.3 |
| BiGGen-Bench | 3.49 | Average Score (1-5) | 60.8 |
| Open Portuguese LLM - BLUEX | 54.1 | Accuracy (%) | 58.6 |
| Open Portuguese LLM - OAB Exams | 43.74 | Accuracy (%) | 56.5 |
| Open Portuguese LLM - ASSIN2 RTE | 90.22 | Macro F1 (%) | 55.1 |
| RewardBench Safety | 82.43 | Accuracy (%) | 55.1 |
| Open Portuguese LLM - ENEM | 63.26 | Accuracy (%) | 55 |
| LogicKor - Reasoning | 6 | Score (0-10) | 53.8 |
| RewardBench | 74.81 | Score (%) | 48.9 |
| RewardBench Chat Hard | 60.53 | Accuracy (%) | 48.6 |
Interactive version: theaggregate.ai/model?slug=nous-hermes-2-mistral-7b-dpo · How It Works · Data refreshed daily, snapshot 2026-10-20.