Codestral Embed
Codestral Embed is Mistral's first embedding model specialized for code, outperforming general-purpose and competing code embedding models on real-world retrieval benchmarks.
import { embed } from 'ai';
const result = await embed({ model: 'mistral/codestral-embed', value: 'Sunny day at the beach',})Providers
Route requests across multiple providers. Copy a provider slug to set your preference. Visit the docs for more info. Using a provider means you agree to their terms, listed under Legal.
More models by Mistral
| Model |
|---|
| 256K | 0.9s | 8tps | $0.10/M | $0.30/M | — |  | 12/09/2025 | |||||
| 256K | 0.4s | 71tps | $0.50/M | $1.50/M | — | | 12/02/2025 | |||||
| 256K | 0.3s | 90tps | $0.20/M | $0.20/M | — | | 12/02/2025 | |||||
| 128K | 0.3s | $0.10/M | $0.10/M | — | | 10/16/2024 | ||||||
| 32K | 0.3s | 182tps | $0.10/M | $0.30/M | — | | 09/17/2024 | |||||
$0.10/M | — | | 12/11/2023 |
About Codestral Embed
Released May 28, 2025, Codestral Embed is Mistral's first embedding model purpose-built for code. Codestral Embed achieves an 85% average score on code retrieval benchmarks, outperforming Voyage Code 3, Cohere Embed v4.0, and OpenAI's large embedding model on evaluations derived from real-world code data.
Codestral Embed supports variable dimensions with ordered relevance. You can truncate to the first n dimensions and still get ranked embeddings. Mistral's benchmarks show retrieval performance even at 256 dimensions with int8 precision, enabling index size reduction without proportional quality loss.
The context window is 0 tokens. For repositories with large files, Mistral recommends chunking at 3,000 characters with 1,000-character overlap. This balances retrieval recall against chunk boundary artifacts.
What To Consider When Choosing a Provider
- Configuration: Codestral Embed supports variable embedding dimensions, letting you tune the size-versus-quality tradeoff to match your vector store's cost and latency constraints.
- Zero Data Retention: AI Gateway supports Zero Data Retention for this model via direct gateway requests (BYOK is not included). To configure this, check the documentation.
- Authentication: AI Gateway authenticates requests using an API key or OIDC token. You do not need to manage provider credentials directly.
When to Use Codestral Embed
Best for
- RAG pipelines for coding agents: Building pipelines that retrieve relevant code snippets
- Semantic code search: Large repositories where keyword search is insufficient
- Duplicate code detection: Similarity analysis across codebases using Codestral Embed's 85% average on code retrieval benchmarks
- Code clustering: For analytics, refactoring identification, or repository organization
- Large-scale indexing pipelines: Workloads where embedding cost is a primary concern at millions of documents
Consider alternatives when
- General text documentation: You need to embed prose rather than source code (consider Mistral Embed)
- Natural language retrieval: Your workload is primarily over descriptions of code rather than code itself
- Generation alongside embedding: You require both capabilities in a single model call
Conclusion
Codestral Embed fills a gap that general-purpose embedding models leave open. Code has structural patterns, syntax, and semantics that differ from prose, and a model trained on real-world code data retrieves it more accurately. For teams building coding agents, code search, or repository analytics, Codestral Embed is more precise than adapting a text embedding model.