N
Nous: Hermes 4 405B
● Excellent8.4
/ 10
Hermes 4 is a large-scale reasoning model built on Meta-Llama-3.1-405B and released by Nous Research. It introduces a hybrid reasoning mode, where the model can choose to deliberate internally with...
Specifications
| Attribute | Value |
|---|---|
| Lab | Nousresearch |
| Tags | Open Weight Reasoning |
| Overall Score | 8.4/10 |
| Release Date | 2025-08 |
| Context Window | 131,072 tokens |
| Input Price / 1M | $1.00 |
| Output Price / 1M | $3.00 |
| Input Modalities | Text |
| Output Modalities | Text |
Strengths
- Hybrid reasoning mode auto-decides when to think deeply
- Strong instruction following and tool use
- Open-weight on Llama 3.1 405B base
- Uncensored — no refusal for controversial topics
Weaknesses
- Based on older Llama 3.1 architecture
- Very large model — slow and expensive to run
- Less competitive on raw benchmarks vs latest models
Best For
- Complex multi-step reasoning
- Creative and uncensored writing
- Research and experimentation
- Tool-use and agentic workflows
Sources & Further Reading
Related Models
Scores are aggregated from public benchmarks (MMLU, HumanEval, MATH, GSM8K, LMSYS) and normalized to a 1–10 scale. Methodology →