Introducing Mistral Small 4

2026-03-16 · Source: Simon Willison's Weblog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, quick

Summary

Mistral has released Mistral Small 4, an Apache 2 licensed 119B parameter Mixture-of-Experts model with 6B active parameters. This new model unifies the capabilities of previous flagship models: Magistral for reasoning, Pixtral for multimodal tasks, and Devstral for agentic coding into a single, versatile offering. It supports a `reasoning_effort` parameter with "none" or "high" settings, where "high" provides verbosity equivalent to prior Magistral models. The model is available on Hugging Face as a 242GB download. Additionally, Mistral introduced Leanstral, an open-weight model specifically fine-tuned for generating code in the Lean 4 formally verifiable programming language.

Key takeaway

For AI Architects evaluating new foundation models, Mistral Small 4 offers a compelling option by consolidating diverse capabilities into one model. This unification could simplify deployment and management compared to using separate specialized models. Monitor the Mistral API for the `reasoning_effort` parameter to fine-tune performance for specific reasoning-intensive applications.

Key insights

Mistral Small 4 unifies multiple specialized AI capabilities into a single, versatile model.

Principles

• Model unification enhances versatility.
• Targeted fine-tuning improves language-specific code generation.

In practice

• Explore Mistral Small 4 for combined reasoning, multimodal, and agentic tasks.
• Consider Leanstral for Lean 4 code generation.

Topics

• Mistral Small 4
• Mixture-of-Experts
• Multimodal AI
• Agentic Coding
• Lean 4

Code references

• simonw/llm-mistral

Best for: AI Architect, NLP Engineer, Computer Vision Engineer, AI Engineer, Machine Learning Engineer, Prompt Engineer

Related on AIssential

• Mistral Small 4: The One Model That Codes, Reasons, and Chats — Mistral Small 4 unifies chat, reasoning, and coding via MoE architecture for efficient, multimodal AI.
• Mistral AI Releases Mistral Small 4: A 119B-Parameter MoE Model that Unifies Instruct, Reasoning, and Multimodal Workloads — Mistral Small 4 unifies diverse AI workloads into a single, efficient MoE model.
• LWiAI Podcast #238 - GPT 5.4 mini, OpenAI Pivot, Mamba 3, Attention Residuals — AI development is rapidly advancing across models, applications, and safety, with a clear industry shift towards agentic AI and enterprise solutions.
• Together AI expands fine-tuning service with tool calling, reasoning, and vision support — Together AI's fine-tuning service now natively supports complex AI workflows like tool calling, reasoning, and VLM training.
• Solutions for any use case | Mistral - Mistral AI — Mistral AI delivers customizable, secure, and scalable AI solutions tailored for diverse enterprise use cases and industry-specific applications.
• Mistral Small 4 in 8 mins! — Mistral Small 4 unifies reasoning, coding, and multimodal capabilities for enterprise applications under an Apache 2.0 license.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.