Devstral Small 1.0

Visit Website
Devstral Small 1.0

Description

️ 🖼Tool name:
Devstral Small 1.0

🔖 Tool Category:
Agentic Code Model
It falls under the category Programming & Development / Generative AI Models.

️ ✏What does this tool offer?

  • Mistral AI' s open-source artificial intelligence model designed specifically for programming tasks.

  • It can analyze code, modify it, and create new solutions by understanding the entire project structure.

  • Supports handling multiple files within a project and executing commands such as code testingand performance optimization.

  • Works with a wide range of languages and tools within development environments.

  • Features a context capacity of up to 128,000 tokens, allowing it to understand huge projects in one go.

  • It can run natively on GPUs like the RTX 4090 or powerful Macs.

  • Licensed under Apache 2.0 (fully open source).

What does the tool actually deliver based on user experience and official news?

  • Outperformed the SWE-Bench Verified tests by 46.8%, the highest performance among open source models in the programming space.

  • When linked with OpenHands Scaffold, it can perform automated operations such as modifying files and running commands within the project.

  • It is recommended to use it with vLLM, Mistral-Inference, or Transformers engines to run it efficiently.

  • Users have noted its high efficiency in code generation and bug analysis within real projects.

🤖 Does it include automation?
Yes - includes extensive automation within development environments:

  • It acts as an "intelligent agent" capable of reading code, modifying it, and executing software tools.

  • It can perform automatic error analysisand apply automatic fixes.

  • It only needs to set up the right environment (scaffold/server) to work at its full potential.

💰 Pricing model:
Open source - no subscription or fees required.
The cost depends only on computational resources (GPU or local server).

🆓 F ree plan details:

  • The model itself is completely free under the Apache 2.0 license.

  • It can be freely downloaded and run on any server or device.

  • No restrictions on non-commercial or commercial use.

💳 Paid plan details:

  • There are no official paid plans.

  • Mistral may later offer Enterprise versions with premium hosting or technical support.

🧭 Method of access:

  • Via the Hugging Face repository:

  • Can be run using tools such as:
    - vLLM
    - Mistral-Inference
    - Transformers
    - LMStudio
    - Ollama

  • Integrated with the OpenHands interface for automated interaction with software projects.

🔗 Trial link:
https://huggingface.co/mistralai/Devstral-Small-2505

Pricing Details

Mistral Medium 3 is completely open source and free, giving users the freedom to modify and use it under the Apache 2.0 license, which allows commercial and non-commercial use without restrictions. However, it is typically run on the Hugging Face platform, which provides a robust environment for hosting and running templates with a range of subscription plans designed to expand the computational and storage capabilities of developers and enterprises. In the Free Plan, public models can be accessed for free, with limited storage and virtualization space, and the model can be downloaded and run locally at no charge. Paid plans start with the PRO plan at $9/month for individual developers and include 10x more private storage, 20x more inference credit, higher priority in ZeroGPU queues, access to Spaces Dev Mode, publishing articles and viewing private data, and a Pro badge on the account. The TEAM plan, priced at $20 per user per month, is for teams and small businesses, and includes all the benefits of PRO plus SSO and SAML support, data location selection (Storage Regions), detailed activity audits (Audit Logs), granular permission control (Resource Groups), repository usage analytics, and centralized authentication and token management policies. The ENTERPRISE plan is geared towards larger organizations and starts at $50 per user per month, and includes all Team features with Custom Onboarding, advanced enterprise support, customized security and data options, the details of which are tailored in direct coordination with the sales team. With this combination of openness and flexibility, developers can run Mistral Medium 3 freely locally, or via Hugging Face in a professional, scalable environment based on their technical and organizational needs.