Models

← All models

Codestral 25.01

Mistral AI · 2025-01-13

Codestral 25.01 (`codestral-2501`) is Mistral AI's closed-weight January 2025 update to its specialized coding model, with a more efficient architecture and improved tokenizer that Mistral said generated and completed code about twice as fast as the original. It is optimized for low-latency, high-frequency code generation, fill-in-the-middle completion, code correction, and test generation. Mistral's launch benchmark labels the context as 256K, while the current retired model card labels it 128K and lists a November 30, 2025 retirement date; Mistral did not disclose an exact parameter count.

Benchmark scores