Source URL: https://simonwillison.net/2025/Jan/30/mistral-small-3/#atom-everything
Source: Simon Willison’s Weblog
Title: Mistral Small 3
Feedly Summary: Mistral Small 3
First model release of 2025 for French AI lab Mistral, who describe Mistral Small 3 as “a latency-optimized 24B-parameter model released under the Apache 2.0 license."
More notably, they claim the following:
Mistral Small 3 is competitive with larger models such as Llama 3.3 70B or Qwen 32B, and is an excellent open replacement for opaque proprietary models like GPT4o-mini. Mistral Small 3 is on par with Llama 3.3 70B instruct, while being more than 3x faster on the same hardware.
Llama 3.3 70B and Qwen 32B are two of my favourite models to run on my laptop – that ~20GB size turns out to be a great trade-off between memory usage and model utility. It’s exciting to see a new entrant into that weight class.
The license is important: previous Mistral Small models used their Mistral Research License, which prohibited commercial deployments unless you negotiate a commercial license with them. They appear to be moving away from that, at least for their core models:
We’re renewing our commitment to using Apache 2.0 license for our general purpose models, as we progressively move away from MRL-licensed models. As with Mistral Small 3, model weights will be available to download and deploy locally, and free to modify and use in any capacity. […] Enterprises and developers that need specialized capabilities (increased speed and context, domain specific knowledge, task-specific models like code completion) can count on additional commercial models complementing what we contribute to the community.
Despite being called Mistral Small 3, this appears to be the fourth release of a model under that label. The Mistral API calls this one mistral-small-2501 – previous model IDs were mistral-small-2312, mistral-small-2402 and mistral-small-2409.
I’ve updated the llm-mistral plugin for talking directly to Mistral’s La Plateforme API:
llm install -U llm-mistral
llm keys set mistral
# Paste key here
llm -m mistral/mistral-small-latest "tell me a joke about a badger and a puffin"
Sure, here’s a light-hearted joke for you:
Why did the badger bring a puffin to the party?
Because he heard puffins make great party ‘Puffins’!
(That’s a play on the word "puffins" and the phrase "party people.")
Mistral also ensured that the new model was available on Ollama in time for their release announcement.
You can pull the model like this (fetching 14GB):
ollama run mistral-small:24b
The llm-ollama plugin will then let you prompt it like so:
llm install llm-ollama
llm -m mistral-small:24b "say hi"
Tags: open-source, mistral, llm, ollama, generative-ai, ai, llms
AI Summary and Description: Yes
Summary: The text discusses the release of Mistral Small 3 by the French AI lab Mistral, highlighting its competitive performance against larger models and its licensing shift to Apache 2.0. This move towards open-source accessibility could significantly impact the AI landscape, particularly for enterprises and developers.
Detailed Description: The text provides an overview of Mistral Small 3, a new AI model developed by the French lab Mistral. Here are the major points discussed:
– **Model Performance**:
– Mistral Small 3 is identified as a “latency-optimized” 24 billion parameter model.
– It competes effectively with larger models such as Llama 3.3 (70 billion parameters) and Qwen 32B, claiming performance parity while being significantly faster (over three times faster) on the same hardware.
– **Licensing Update**:
– The model is released under the Apache 2.0 license, a shift from the previous Mistral Research License that restricted commercial use.
– This will allow users to download, modify, and deploy the model freely, which is a substantial change that supports broader accessibility and innovation.
– **Market Positioning**:
– Mistral’s strategy seems to favor open-source initiatives, positioning Mistral Small 3 as an alternative to proprietary models like GPT4o-mini, which are more opaque and often come with limitations on usage.
– There is an emphasis on providing enterprises and developers access to specialized capabilities through additional commercial models, alongside their open-source offerings.
– **Model Identification**:
– While designated as Mistral Small 3, this is noted to be the fourth variation of models under that name, indicating ongoing iterations and improvements in their offering.
– **Integration Capability**:
– Compatibility with the Mistral API and other tools (like llm-mistral and llm-ollama plugins) is mentioned, showing the ease with which users can interact with the new model in practical settings.
The proactive move towards open licensing and performance optimization reflects a growing trend in the AI industry—promoting transparency and accessibility while maintaining competitive edge. This is particularly relevant for security and compliance professionals as the use and deployment of AI tools become more entwined with data governance and regulation adherence. Open-source models can significantly influence compliance frameworks by providing clear guidelines on usage and modification, which is crucial in sectors requiring stringent security measures.