unsloth/Devstral-Small-2507

Hugging Face
TEXT GENERATIONConcurrency Cost:2Model Size:24BQuant:FP8Ctx Length:32kPublished:Jul 10, 2025License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

Devstral-Small-2507 is a 24 billion parameter agentic language model developed by Mistral AI and All Hands AI, fine-tuned from Mistral-Small-3.1. It is specifically designed for software engineering tasks, excelling at using tools to explore codebases and edit multiple files. With a 128k token context window, it achieves a 53.6% score on SWE-Bench Verified, positioning it as a leading open-source model for agentic coding.

Loading preview...

Devstral-Small-2507: Agentic LLM for Software Engineering

Devstral-Small-2507 is a 24 billion parameter agentic language model, a collaboration between Mistral AI and All Hands AI. Fine-tuned from Mistral-Small-3.1, this model is engineered to excel in software engineering tasks, particularly those involving tool use, codebase exploration, and multi-file editing. It features a substantial 128k token context window and supports Mistral's function calling format.

Key Capabilities

  • Agentic Coding: Optimized for complex software engineering workflows, enabling agents to interact with codebases and perform edits.
  • High Performance on SWE-Bench: Achieves a 53.6% score on SWE-Bench Verified, outperforming previous versions and other state-of-the-art models in its class.
  • Lightweight Deployment: Despite its capabilities, the 24B parameter model is designed to run on consumer-grade hardware like an RTX 4090 or a Mac with 32GB RAM, facilitating local and on-device inference.
  • Apache 2.0 License: Offers broad commercial and non-commercial usage and modification rights.
  • Tool Use: Supports Mistral's function calling format, enhancing its ability to integrate with external tools and environments.

Good For

  • Software Engineering Agents: Ideal for building autonomous agents that can understand, analyze, and modify code.
  • Codebase Exploration and Editing: Excels at tasks requiring deep interaction with code, such as identifying issues, refactoring, or implementing new features across multiple files.
  • Local Development Environments: Its efficient design allows for deployment on local machines, making it suitable for developers who need powerful coding assistance without cloud dependency.
  • Benchmarking and Research: Provides a strong baseline for research into agentic LLMs and software development automation.

Finetunes
4 models