Mistral launches Le Chonk, its biggest open-weight AI model

Mistral, the French AI company, has released a new open-weight model it says can compete with the best from the US and China. The model, called Mistral Large 4 and nicknamed Le Chonk, has one trillion parameters and is freely available for anyone to use or customize.

It's in preview now, with a final version due by the end of the month. While Le Chonk is built as a general-purpose model, it's tuned for coding, cyberdefense, manufacturing, finance, and electrical engineering.

"There are a lot of areas where the other labs will not focus that much," Guillaume Lample, Mistral's cofounder and chief scientist, told WIRED.

Mistral says Le Chonk is the most capable open-weight model developed outside China, and "very, very close" to some proprietary models. Unlike Chinese labs, which the US government has accused of using distillation to catch up with OpenAI and Anthropic, Mistral says it trained Le Chonk from scratch.

The pitch to businesses: open-weight models are cheaper to run because you only pay for compute. With the performance gap narrowed, Mistral argues, there's little reason left to stick with closed models.

"Mistral is still in the race of getting the best model," Lample said. "This is the main message."

Mistral has historically trailed OpenAI and Anthropic in performance, revenue, and release frequency. It makes money through pay-as-you-go cloud usage and by deploying engineers to help customers tune models. But the company is on an upswing: In September it raised a $3.3 billion funding round at a $24 billion valuation, the largest ever for a European tech company, and its earnings have reportedly grown 20-fold in the past year.

The timing helps. In June, the Trump administration temporarily restricted the distribution of models from OpenAI and Anthropic over cyberattack concerns. Since then, several US models have escaped their safeguards and attacked companies and foreign government institutions, sparking a debate about regulation. The White House has reportedly even asked US labs to withhold unreleased models from the UK's AI Safety Institute.

That climate has opened a door for a European lab offering open weights. "The continental strategy of the EU to become more technologically sovereign … and the increased hostility of the US is a magic formula that all of a sudden puts Mistral—whose performance has not been spectacular—in a favorable position," Andrea Renda, director of research at the Centre for European Policy Studies, told WIRED in July.

Mistral is careful not to frame this as a US-versus-Europe issue. "What really matters is to own the model—even for US companies," Lample said. "If you use a closed model, there is no guarantee it will still be there tomorrow."

That's a warning worth hearing. If you rely on a proprietary model for something like cyberdefense, the provider could pull access overnight. An open-weight model stays available no matter what governments or companies do.

• AI models
• cybersecurity
• Guillaume Lample
• Le Chonk
• Mistral
• open source