tech
The Platform
Mistral AI brings the strongest open generative models to the developers, along with efficient ways to deploy and customize them for production.
TL;DR
- Mistral AI's new platform offers beta access to generative text and embedding endpoints.
- Three chat endpoints (mistral-tiny, mistral-small, mistral-medium) vary in performance, price, and supported models.
- Mistral-tiny uses Mistral 7B Instruct v0.2 (English only), Mistral-small uses Mixtral 8x7B (multilingual), and Mistral-medium uses a prototype model (multilingual).
- An embedding endpoint, Mistral-embed, is available with a 1024 embedding dimension.
- The API specifications follow a popular chat interface, with Python and Javascript client libraries available.
- Users can provide system prompts for higher moderation levels.
- NVIDIA is acknowledged for their support in integrating TensorRT-LLM and Triton.