Mistral launches Mistral 3, a family of 10 models under the Apache 2.0 license, including its new flagship Mistral Large 3 and nine smaller Ministral 3 models
Mistral AI, Europe's most prominent artificial intelligence startup, is releasing its most ambitious product suite to date …
VentureBeatMichael Nuñez
Context & Ripple Effects
Mistral has been building a segmented model lineup: code and math models in 2024, smaller on-device-oriented Ministraux models, and a latency-focused Small 3 release. It then added Apache-licensed reasoning capability alongside a separately previewed medium model.
Mistral 3 consolidates that release pattern into a 10-model family, pairing a new flagship with nine smaller variants under Apache 2.0. The move matters because it extends the company’s open distribution approach from individual releases to a broader portfolio.
First-order effects
Developers and enterprise buyers gain an Apache 2.0-licensed choice across a flagship model and nine smaller Ministral models, rather than selecting from isolated Mistral releases.
Mistral can position Large 3 and the Ministral range as a coordinated portfolio, while the license permits downstream use and modification subject to its terms.
Second-order effects
Organizations evaluating Mistral against proprietary and more narrowly distributed models can compare deployment fit across model sizes without changing vendors or licensing frameworks.
The expanded range raises the value of Mistral’s earlier small-model, on-device-oriented line: smaller variants can serve different latency and deployment needs while remaining within the same family.
Third-order effects
If more frontier-capable model families are released under permissive terms, differentiation shifts further from basic model access toward deployment, integration, and operational model selection.
A growing portfolio approach could make AI buying more explicitly tiered: buyers may match workloads to model size and capability rather than standardize on one general-purpose model.
The trend: This is part of the shift from standalone model launches toward licensed, multi-tier model portfolios designed for different deployment and procurement needs.
It's a perfect day to announce that I've joined Mistral as an AI scientist, when our new flagship model has arrived :). Obviously, I did not contribute to this one, but I have high hopes about the next one :). I am very excited about this opportunity for a few reasons. On the
Massive move from @MistralAI dropping the Mistral 3 family as Open Source today. Finally, capable multimodal models (vision included) that we can run on-prem. The best part? Ministral 14B runs comfortably on a single 24GB GPU. 🤯This is the new sweet spot for fast, local agents. […
Mistral 3 is now available on Ollama v0.13.1 (currently in pre-release on GitHub). 14B: ollama run ministral-3:14b 8B: ollama run ministral-3:8b 3B: ollama run ministral-3:3b Please update to the latest Ollama. [image]
🎉 Excited to introduce @MistralAI Ministral 3 (3B, 8B, and 14B) and Mistral Large 3 (sparse MOE trained w/ 41B active and 675B total parameters) - All open weights under the Apache 2.0 license! Details in 🧵: [image]
NEW: @MistralAI releases Mistral 3, a family of multimodal models, including three start-of-the-art dense models (3B, 8B, and 14B) and Mistral Large 3 (675B, 41B active). All Apache 2.0! 🤗 Surprisingly, the 3B is small enough to run 100% locally in your browser on WebGPU! 🤯 [vide…
i guess mistral large 3 is the SOTA open source model for it's param size now. better instruction following than deepseek 3.1 / kimi k2 (non thinking variants). they are also releasing a nvfp4 checkpoint that can run on a single A100 or 8xh100 node [image]
Mistral Large 3 was a stealth model called Spectre. All models are released under the Apache 2.0 license. This model is optimized for software development tasks. Impressive! [image]
Mistral just dropped Mistral Large 3 on Hugging Face! Their new state-of-the-art multimodal, Mixture-of-Experts model boasts 41B active params, 675B total, and a massive 256k context window for frontier performance. [image]