The Mistral 3 models include three small, dense Ministral models (14B, 8B and 3B) and the company’s most capable model to date, Mistral Large 3, the company said in a Tuesday press release.
Mistral Large 3 is trained from scratch on 3,000 Nvidia H200 GPUs and is Mistral’s first mixture-of-experts model since the Mixtral series, according to the release.
“After post-training, the model achieves parity with the best instruction-tuned open-weight models on the market on general prompts, while also demonstrating image understanding and best-in-class performance on multilingual conversations (i.e., non-English/Chinese),” the release said.
The Ministral models are available not only in three model sizes but also in three variants: base, instruct and reasoning, per the release. These models are designed for edge and local use cases.
“In real-world use cases, both the number of generated tokens and model size matter equally,” the release said. “The Ministral instruct models match or exceed the performance of comparable models while often producing an order of magnitude fewer tokens.”
Together, the new Mistral 3 models feature frontier intelligence, multimodal flexibility and unmatched customization, according to the release.
“Whether you’re deploying edge-optimized solutions with Ministral 3 or pushing the boundaries of reasoning with Mistral Large 3, this release puts state-of-the-art AI directly into your hands,” the release said.
Mistral raised €1.7 billion (about $2 billion) in a Series C funding round in September that valued the company at €11.7 billion (about $13.6 billion).
The company said at the time that it would continue to develop customized frontier AI solutions designed to solve engineering and industrial problems.
The funding round was led by semiconductor equipment manufacturer ASML.
“The collaboration between Mistral AI and ASML aims to generate clear benefits for ASML customers through innovative products and solutions enabled by AI, and will offer potential for joint research to address future opportunities,” ASML CEO Christophe Fouquet said at the time in a press release.
In October, Mistral introduced a production AI platform called Mistral AI Studio that helps enterprise teams build, evaluate and run AI in production.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
