Codestral

Frontier

Mistral AI•Released on 2025-01-14

Mistral's dedicated coding model. SOTA for Fill-in-the-Middle (FIM) use cases. 2x faster code generation than original. 256K context for large codebases. Excellent for autocomplete and code completion.

77
Overall Score

Core Specs

256K
Context Window
33K
Max Output
✗
Reasoning
✗
Open Source
Multimodal Support
text

Scenario Scores

Coding
#8 / 2488
Writing
—
Translation
—
Data Analysis
#24 / 2465
Conversation
—
Image Gen
—

User Feedback Highlights

Based on community feedback. Hover to see original reviews.

− Code-only, not general purpose− Not open source (unlike Mistral Large)− Less capable for complex reasoning+ 2x faster than original Codestral+ Very affordable ($0.3/$0.9)+ Optimized for IDE integration+ SOTA for FIM/autocomplete+ 256K context for large codebases− No multimodal support
Sentiment:👍 75%😐 18%👎 7%

Pros & Cons

Pros

  • +SOTA for FIM/autocomplete
  • +2x faster than original Codestral
  • +256K context for large codebases
  • +Very affordable ($0.3/$0.9)
  • +Optimized for IDE integration

Cons

  • −Code-only, not general purpose
  • −No multimodal support
  • −Not open source (unlike Mistral Large)
  • −Less capable for complex reasoning

Reliability

Incidents (30d)0
Mistral API 托管

Pricing

Input (per 1M tokens)$0.30
Output (per 1M tokens)$0.90
Free trial available
Updated on 2026-03-07

Compare with Others

Codestral vs GPT-5.4 →Codestral vs GPT-5.4 Thinking →Codestral vs Claude Opus 4.6 →

Benchmarks

humanEval86.6%
mbpp78.2%