TL;DR: Mistral published the weights for Mistral Large 4, nicknamed Le Chonk, on Hugging Face, so the flagship model can now be self-hosted instead of only called via API.
The Catalyst
The release is a weights drop on Hugging Face for Mistral Large 4, referred to as Le Chonk. The post gives no parameter count, license, context length or benchmark numbers.
The Blast Radius
Teams paying per token for Mistral's hosted flagship now have the option of running the same model in their own infrastructure, which shifts the cost question from token price to GPU memory and throughput. Check the model card for license terms and the weight file sizes to see what hardware class you need before planning a migration.
Source: x_com