Last updated: 16/04/2025

MistralOfficial Docs

Codestral

codestral-2405

deprecated

Codestral

Codestral is a cutting-edge language model for coding, specializing in low-latency, high-frequency tasks such as fill-in-the-middle (FIM), code correction, and test generation.

Supports a 32K token context window. Handles Text, Image, Video, Audio, Transcription, Text-to-Speech inputs and outputs. Supports fine-tuning for custom applications. Supports tool use for advanced automation. Capable of generating structured output formats.

Additional Information

Notes

The latest version (25.01) was released in January 2025. The model has a context length of 256k tokens. It's available via API as 'codestral-latest' which currently points to 'codestral-2501'. The original version (24.05) has been marked as legacy and deprecated.

Model Timeline

Launch Date

5/1/2024

Last Updated

1/1/2025

Marked Legacy

12/2/2024

Marked Deprecated

12/2/2024

Marked Expired

3/30/2025

Capabilities

Text

Input Pricing

$0.50/ MTok

Context: 32,768 tokens

Output Pricing

$1.50/ MTok

Max tokens: 32,768

Embeddings

Embeddings Pricing

$0.10/1k tokens

Additional Model Information

Tool Use

Yes

Structured Output

Yes

Reasoning

Yes

Flatten your repo for AI in seconds

Flatten repos. Prompt faster. One click → one GPT-ready file

Free Online & Desktop