Mistral: Devstral Medium

Devstral Medium is a cutting-edge AI model developed jointly by Mistral AI and All Hands AI, specifically engineered for high-performance code generation and advanced agentic reasoning. This model represents a significant leap from its predecessor, Devstral Small, achieving an impressive 61.6% on SWE-Bench Verified. This benchmark places it ahead of industry leaders like Gemini 2.5 Pro and GPT-4.1 in critical code-related tasks, all while offering a fraction of the cost. Designed for versatility, Devstral Medium excels in generalizing across various prompt styles and is highly effective for tool use within code agents and frameworks. It supports a robust context window of 131K tokens and can generate outputs up to 4K tokens. Available exclusively via API, it supports enterprise deployment on private infrastructure and offers optional fine-tuning capabilities. Pricing is competitive at $0.40 per 1M input tokens and $2.00 per 1M output tokens, making it a cost-effective solution for advanced coding needs. It is best suited for chat, code generation, and translation tasks. This model is part of Multi AI's PRO access tier, providing premium access to its advanced features. While powerful in code and text, it does not support image generation or internet access.