Devstral is an advanced AI model developed by Mistral AI, designed to enhance software engineering tasks through its agentic coding capabilities. Released in July 2025, Devstral offers two versions: Devstral Small 1.1 and Devstral Medium, each tailored to meet different performance and deployment needs.
Key Features
- Performance: Devstral Small 1.1 achieves a score of 53.6% on SWE-Bench Verified, setting a new state-of-the-art for open models without test-time scaling. Devstral Medium surpasses this with a score of 61.6% on the same benchmark.
- Versatility: Both versions support integration with various agentic scaffolds, including OpenHands, and are compatible with Mistral function calling and XML formats, allowing for adaptability across different coding environments.
- Deployment Options: Devstral Small 1.1 is available under the Apache 2.0 license, enabling local deployment and customization. Devstral Medium is accessible via Mistral’s API and can be deployed on private infrastructure, offering enhanced data privacy and control.
Who is it for?
Devstral is suitable for software developers, engineers, and organizations seeking to integrate advanced AI capabilities into their coding workflows. Its open-source nature and flexible deployment options make it appealing for both individual developers and enterprise teams aiming to enhance their software development processes.
Pricing
Devstral is available through Mistral’s API under the following pricing structure:
- Devstral Small 1.1: $0.10 per million input tokens and $0.30 per million output tokens.
- Devstral Medium: $0.40 per million input tokens and $2.00 per million output tokens.
For those interested in local deployment, Devstral Small 1.1 is released under the Apache 2.0 license, allowing for free use and modification.
Final Thoughts
Devstral represents a significant advancement in AI-driven software engineering tools, offering high performance and flexibility. Its open-source licensing and versatile deployment options make it a compelling choice for developers and organizations looking to integrate AI into their coding practices. Prospective users should consider their specific requirements and evaluate both versions to determine the best fit for their needs.
Visit mistral.ai/news/devstral-2507 for more.
