Mistral is a family of language models from Mistral AI, several released as open weights. The line includes dense models and mixture-of-experts models for text generation and reasoning.
Semantic Classification
Content
- Mistral models are transformer language models that gained attention for releasing competitive open-weight versions that can be run locally or self-hosted. The range spans smaller dense models and larger mixture-of-experts designs.
- The models are used for text generation, summarisation and tool-calling agents, and are available both as downloadable weights and through a hosted API. Open releases have made them a common base for fine-tuning.