Codestral Embed 2505
by Mistral AI
Embedding
Codestral Embed 2505 is a purpose-built embedding model for code, designed to generate embeddings with varying dimensions and precisions to balance retrieval quality and storage efficiency. It is optimized for large-scale developer workflows, enabling fast and precise semantic code search and retrieval across massive codebases. Compared to alternatives such as Voyage Code 3, Cohere Embed v4.0, and OpenAI’s large embedding model, Codestral-embed-2505 delivers superior performance in both accuracy and speed. This makes it particularly effective for powering advanced developer tools, code intelligence systems, and retrieval-augmented generation pipelines in coding environments.
Provider | Context Size | Throughput | Latency | Input Cost | Output Cost |
|---|---|---|---|---|---|
Usage
Generate your API key and query the model through the OpenAI-compatible interface. The preference parameter allows you to define the routing strategy. For more details, see the documentation.