Mistral AI SAS releases a preview of “Mistral Large 4,” its most capable large language model to date, and makes it available through its cloud platform and API. Multiple outlets report that the model is offered as a public preview initially, with developers able to access it starting today.
Both sources describe the model as a 1-trillion-parameter system using a mixture-of-experts architecture. They also agree that Mistral plans to release the model’s weights later, enabling users to download and run it on their own infrastructure. The Next Web notes that this weight release is scheduled for October, while SiliconANGLE states the weights will be published later this month.
Across the coverage, the main differences are in emphasis: one outlet highlights the open-source framing and the company’s broader AI roadmap, while the other focuses on the “open-weight” nature of the release and the practical developer pathway via Mistral’s API. Both portray the launch as an expansion of access to a large, open model outside China, according to Mistral.
