Nous Hermes 2 Mixtral 8x7B SFT
The SFT release adapts Mixtral 8x7B through supervised instruction tuning. Its training used more than one million examples, mainly generated with GPT-4. It uses ChatML for conversation prompts.
Training and model behavior
This is the supervised-only Mixtral Hermes release. It was trained on more than one million examples. Nous also publishes a version with preference optimization for direct comparison.Nous Hermes 2 Mixtral 8x7B SFT model card
Working with this model
Practical guidance. Preserve the model's ChatML roles and stop tokens. Compare style and factual accuracy separately. Older benchmark positions describe the release period rather than the current model market.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | Downloadable |
| Model identifiers | NousResearch/Nous-Hermes-2-Mixtral-8x7B-SFT |
| Context window | 32,768 tokens Configured position limit; practical serving limits depend on the runtime. Nous Hermes 2 Mixtral 8x7B SFT configuration |
| Base model | mistralai/Mixtral-8x7B-v0.1Nous Hermes 2 Mixtral 8x7B SFT model card |
| Repository license label | apache-2.0Nous Hermes 2 Mixtral 8x7B SFT model card |
Related models
- Nous Hermes 2 SOLAR 10.7B
Nous Research · Hermes 2
An earlier instruction-tuned Hermes model with downloadable weights.
Launch date January 2024Downloadable - Nous Hermes 2 Yi 34B
Nous Research · Hermes 2
An earlier instruction-tuned Hermes model with downloadable weights.
Launch date Downloadable - Nous Hermes 2 Llama 2 70B
Nous Research · Hermes 2
An earlier instruction-tuned Hermes model with downloadable weights.
Launch date not confirmedDownloadable