Mistral AI previews Large 4, with open weights planned for late October
The AMW Read
Large 4 meaningfully updates an established foundation-model player's offering, with a deliberate open-weight release strategy and segment-level implications for enterprise model selection.
Mistral AI previews Large 4, with open weights planned for late October
Mistral AI launched the public preview of Mistral Large 4 on October 6, with downloadable weights scheduled for the end of the month. The French lab describes a sparse mixture-of-experts model with 1 trillion total parameters and 49 billion active parameters. It accepts text and images, supports more than 160 languages, and offers a 524,288-token context window. Developers can access the preview through Mistral Studio. Reported API pricing is $1.36 per million input tokens and $4.18 per million output tokens.
The release positions Mistral around two competitive levers in foundation models: open-weight availability and European operational control. The company says it trained Large 4 entirely in its own European data centers and operates that infrastructure independently under EU law. That gives enterprise buyers a deployment proposition beyond benchmark rankings, but does not establish frontier parity. The report cites an Artificial Analysis Intelligence Index score of 38 and a 62% result on DeepSWE v1.1, while noting that leading closed models achieve higher software-engineering scores. Its reported strength on Harvey's Legal Agent suggests workload-specific performance deserves closer attention than an overall leadership claim.
Builders should distinguish today's hosted preview from the planned weights release. For teams evaluating European deployment options, the concrete next step is to test Large 4 on their own multilingual, visual, and legal workflows at the stated API prices. Self-hosting decisions should wait for the weights and accompanying release terms; the article provides no license details or serving requirements. Investors should assess whether European control and targeted enterprise performance translate into customer demand despite the reported gap on broader frontier benchmarks.
