Mistral, a well-funded French AI model maker, has released its latest open-source language model, Devstral. The model boasts 24-million parameters and is designed to act as a full software engineering agent, capable of understanding context across files, navigating large codebases, and resolving real-world issues.
Devstral represents the next step in Mistral’s growing portfolio of code-focused models, building upon its earlier success with the Codestral series. The model achieves a score of 46.8% on the SWE-Bench Verified benchmark, outperforming previously released open-source models and closed models like GPT-4.1-mini.
Developed in collaboration with open source startup All Hands AI, Devstral is optimized for integration into agentic frameworks like OpenHands, SWE-Agent, and OpenDevin. The model is freely available under the permissive Apache 2.0 license, allowing developers and organizations to deploy, modify, and commercialize it without restriction.
Mistral’s research scientist, Baptiste Rozière, stated that the goal was to release a model open for the developer and enthusiast community, something they can run locally, privately, and modify as they want. The model is finetuned from Mistral Small 3.1 using reinforcement learning and safety alignment techniques.
Devstral’s compact architecture makes it practical for developers to run locally, even on a single RTX 4090 GPU or a Mac with 32GB of RAM. Its Apache 2.0 license offers a compelling proposition for commercial applications, permitting unrestricted use, adaptation, and distribution – making Devstral a low-friction option for enterprise adoption.
The model features a 128,000 token context window and uses the Tekken tokenizer with a 131,000 vocabulary. It supports deployment through all major open source platforms, including Hugging Face, Ollama, Kaggle, LM Studio, and Unsloth, and works well with libraries such as vLLM, Transformers, and Mistral Inference.
While Devstral is currently released as a research preview, Mistral and All Hands AI are already working on a larger follow-up model with expanded capabilities. With its performance benchmarks, permissive license, and agentic design, Devstral positions itself not just as a code generation tool – but as a foundational model for building autonomous software engineering systems.
Source: https://venturebeat.com/ai/mistral-ai-launches-devstral-powerful-new-open-source-swe-agent-model-that-runs-on-laptops