Hermes 3 - Llama-3.1 70B: benchmark results
Nous Research's Hermes 3 70B, their flagship generalist chat model built on Llama 3.1 70B, with agentic capabilities, roleplay, structured function calling and ChatML-formatted multi-turn conversation. Provider: Nous Research. Released 2024-08-16. Access: Open.
Aggregate Rating 1462 ± 24, rank #927 of 1656 rated models, from 100 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - MuSR | 49.49 | Score | 97.7 |
| AGC-Bench - unfun_corpus | 1.37 | Dataset z-score | 97.4 |
| AGC-Bench - creatset | 1.61 | Dataset z-score | 97 |
| Open LLM Leaderboard - BBH | 67.56 | Score | 96.1 |
| Open LLM Leaderboard - IFEval | 76.61 | Score | 92.3 |
| BenchBench | 84.51 | Aggregate Score (%) | 91.9 |
| AGC-Bench - science_analogies | 1.5 | Dataset z-score | 91.2 |
| Open LLM Leaderboard - GPQA | 36.16 | Score | 90.8 |
| Open CoT - LogiQA | 7.99 | CoT Gain (%) | 88.9 |
| Open CoT - LSAT Logical Reasoning | 18.63 | CoT Gain (%) | 87.8 |
| Open LLM Leaderboard - MMLU-Pro | 47.27 | Score | 86.1 |
| Open CoT - LSAT Analytical Reasoning | 7.39 | CoT Gain (%) | 84.7 |
Interactive version: theaggregate.ai/model?slug=hermes-3-llama-3-1-70b · How It Works · Data refreshed daily, snapshot 2026-10-20.