Local LLM model page

Codestral (22B)

Mistral's first code model. Supports 80+ programming languages. 476K downloads.

Find the best model for my hardware Browse all 183 LLMs

Parameters

22B

Minimum RAM

16 GB

Model size

13 GB

Quantization

Q4_K_M

Can Codestral (22B) run locally?

Codestral (22B) is best suited for mainstream Macs and PCs with 16 GB RAM. LocalClaw recommends Q4_K_M as the default quantization, with at least 16 GB RAM.

Search term for LM Studio or compatible runtimes: codestral-22b-v0.1

Hugging Face repository: lmstudio-community/Codestral-22B-v0.1-GGUF

codepower

Strengths

Mistral's first code model
80+ languages
Strong completion and generation
476K downloads

Limitations

Non-production license
32K context
Not for commercial use without license

Best use cases

Code generation
Code completion
Refactoring
Code review

Benchmarks

Speed: 5/10

Quality: 7/10

Coding: 9/10

Reasoning: 7/10

Technical details

Developer: Mistral AI

License: MNPL (Non-Production License)

Context window: 32,768 tokens

Architecture: Transformer optimized for code

Released: 2024-05