Devstral or Codestral: which Mistral coding model to take
Both come from Mistral, both need a GPU and both are in Ollama, but they operate in different modes. Devstral is built for agentic development: the model reads the repository, edits files and runs commands on its own; Devstral Small 24B under Apache 2.0 fits on a single GPU, while Devstral 2 123B uses a modified MIT that excludes companies with revenue above 20 million dollars a month. Codestral covers more than 80 programming languages, but the open weights of the 22B version fall under the Mistral Non-Production License, meaning research and testing only; the one free option is Mamba-Codestral 7B under Apache 2.0. Devstral is considerably newer: December 2025 against July 2024.
Comparison based on catalog data
| Parameter | Devstral | Codestral |
|---|---|---|
| Category | Code | Code |
| Developer | Mistral AI (with All Hands AI), France | Mistral AI, France |
| Releases | May 2025 – Dec 2025 | May 2024 – Jul 2024 |
| Sizes | 24B – 123B | 7B – 22B |
| Hardware | 1 GPU, Cluster | Laptop, 1 GPU |
| Commercial use | Commercial use with conditions | Non-commercial only |
| License | Small 24B versions: Apache 2.0; Devstral 2 123B: modified MIT (not for companies with revenue above 20 million dollars per month) | Codestral 22B: Mistral Non-Production License (research and testing only); the spin-off Mamba-Codestral 7B: Apache 2.0 |
| Russian | Supported | Not supported |
| Ollama | Yes | Yes |
| Without GPU | No | No |
| Tasks |
|
|
Choose Devstral if
- You want an agent that fixes tracker tickets and edits code itself
- Commercial use is required: Devstral Small 24B is Apache 2.0
- You have a single GPU to host a 24B model
Choose Codestral if
- The scope is evaluation and testing with nothing going to production
- Programming language coverage matters: Codestral spans over 80
- 7B is enough for you: Mamba-Codestral ships under Apache 2.0
Other comparisons
- Codestral or Qwen Coder: which coding model you may ship
- Gemma or Llama: which to choose for business
- Mistral or Llama: which to choose for business
- Mistral or Qwen: which to choose for business
- Gemma or Phi: small models for modest hardware
- Kimi or DeepSeek: large open models for agentic work
- Cohere Command or Llama: which to choose for a knowledge base
- Qwen Coder or DeepSeek-Coder: which coding model to pick
- StarCoder or Code Llama: which base to fine-tune on
- MiniCPM or Gemma: which model to put on the device
- Qwen or GigaChat: which to choose for business
- Qwen or Llama: which to choose for business
Need a model for your task?
An open model can run on your own server: data stays in-house, there is no per-request fee, and the model can be fine-tuned on your documents.
- SelectThe model and size for your task and hardware budget
- DeployOn your server or in a closed network, with an API
- Fine-tuneOn your data, or connect a knowledge base
- IntegrateInto your CRM, ERP, bot, website or team chat