Mistral Embed
Mistral Embed converts sentences and paragraphs into vectors for search, classification and similarity matching. The December 2023 model remains in Mistral's active catalog. Its purpose is to represent text for comparison, rather than produce a written answer.
Text length does not change vector length
The embedding guide returns one 1,024-value vector for each text string, regardless of its length. It describes using distances between vectors to compare meanings. This fixed output size is useful when planning an index, but does not remove the model's input limit.Mistral text embedding API guide
Keep a consistent search representation
Practical guidance. Build a retrieval sample with the kinds of questions that users will ask. Check whether the correct passage appears near the top. If the embedding model changes, rebuild and compare the stored representations rather than assume that vectors from different models share the same meaning.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API |
| Model identifiers | mistral-embed-2312mistral-embed |
| Context window | 8,000 tokens The model page advertises 8k input context. Mistral Embed API model page |
| Model identifiers | mistral-embed-2312; the text embedding guide uses mistral-embedMistral Embed API model pageMistral text embedding API guide |
| Output size | 1,024 dimensions for each input textMistral text embedding API guide |
| Access | General availability Premier API, with batchingMistral Embed API model page |
Related models
- Codestral Embed
Mistral AI ยท Codestral
A code-retrieval embedding model with selectable vector dimensions and precision.
Launch date API