Nous Hermes 13B
Nous Hermes 13B was fine-tuned on more than 300,000 instructions, mostly from synthetic GPT-4 outputs.
Training and model behavior
Teknium and Karan4D led training and data curation, with compute from Redmond AI. The training used 2,000-token sequences. The release uses Alpaca prompts for instructions and optional input.Nous Hermes 13B model card
Working with this model
Practical guidance. Use this release to reproduce early Hermes behavior. Keep training sequence length separate from the configured context limit. Compare answers with sources when evaluating factual tasks.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | June 2023 |
| Access | Downloadable |
| Model identifiers | NousResearch/Nous-Hermes-13b |
| Context window | 2,048 tokens Configured position limit; practical serving limits depend on the runtime. Nous Hermes 13B configuration |
| Base model | Llama 13BNous Hermes 13B model card |
| Repository license label | gplNous Hermes 13B model card |
Related models
- Nous Hermes Llama 2 70B
Nous Research · Nous Hermes
An original-dataset Hermes fine-tune of Llama 2.
Launch date August 2023Downloadable - Nous Hermes Llama 2 13B
Nous Research · Nous Hermes
An original-dataset Hermes fine-tune of Llama 2.
Launch date Downloadable - Nous Hermes Llama 2 7B
Nous Research · Nous Hermes
An original-dataset Hermes fine-tune of Llama 2.
Launch date July 2023Downloadable