Mistral sparse mixture-of-experts model with strong quality and efficient inference.
Specifications
- Provider
- Mistral AI
- Type
- Open-source / open-weight
- Modality
- Text (MoE)
- Category
- Language model
- Context window
- 64K
- Parameters
- 141B total, ~39B active per token (8x22B MoE)
- License
- Apache 2.0
- Released
- April 17, 2024
What it was trained for
General-purpose language and coding tasks as an open sparse mixture-of-experts model.
Best for
- ▸General chat and writing
- ▸Coding tasks
- ▸Multilingual European-language tasks
- ▸Cost-efficient self-hosted deployment
Capabilities
Sparse MoE (8 experts, 2 active per token)Efficient inference relative to total sizeMultilingual supportNative function callingOpen weights (Apache 2.0)
Performance & positioning
A strong open Apache-licensed MoE model offering a favorable balance of quality and inference cost for its size.
More from Mistral AI