Mistral 发布一万亿参数开源权重模型 Mistral Large 4（Le Chonk）

As tensions mount over who gets access to top-end artificial intelligence, French company Mistral has released a new freely-available model that it claims can compete with the very best from the US and China.

The new one trillion-parameter model, Mistral Large 4—nicknamed Le Chonk—can be used and customized by anyone. It’s currently available in preview, with a final version to follow by the end of the month. Though Le Chonk is built to compete with leading general purpose models, it’s optimized specifically for coding and cyberdefense, as well as tasks particular to manufacturing, finance, electrical engineering, and other niches.

“There are a lot of areas where the other labs will not focus that much,” Guillaume Lample, cofounder and chief scientist at Mistral, tells WIRED. “There are so many domains in which you can improve models.”

Mistral presents Le Chonk as by far the most capable open-weight model developed outside of China, and “very, very close” to some proprietary models. Whereas Chinese labs have been accused by the US government of abusing distillation—the training of a smaller model on the outputs of a larger one—to close the performance gap with OpenAI and Anthropic, Mistral claims to have trained its model from scratch.

It’s already considerably cheaper for businesses to run open-weight models, which cost only as much as the compute they consume. By reducing the performance gap on leading proprietary models and providing a competitive alternative to releases from China, Mistral says, Le Chonk will eliminate the few remaining reasons a business might hesitate to choose open source.

“Mistral is still in the race of getting the best model,” Lample says. “This is the main message.”

With less capital and fewer compute resources than OpenAI and Anthropic, Mistral has generally lagged behind on model performance, revenue, and the frequency of releases. While the US labs charge a premium for access to their proprietary, closed-weight models, Mistral generates revenue by charging pay-as-you-go fees for running models through its cloud, and deploying engineers to help customers tune models to their specific needs. Recently, however, the French lab has been on a hot streak: In September, it raised a $3.3 billion funding round at a $24 billion valuation, the largest ever raise by a European tech company. Its earnings have reportedly increased 20-fold in the last year or so.

The upswing for Mistral coincides with growing animosity between the US and its transatlantic allies—over issues ranging from tariff policy, to Greenland, to the policing of American tech firms—and an increasingly fractious debate over who gets access to frontier-grade AI.

In June, the Trump administration placed temporary restrictions on the distribution of models from OpenAI and Anthropic, citing concerns they could be abused to launch sophisticated cyberattacks. Since then, a litany of incidents have come to light where US-made models have broken free of their constraints and attacked companies and some foreign government institutions, leading to weeks of debate over how model releases should be regulated. The White House has reportedly asked the American labs to withhold unreleased models even from the UK’s AI Safety Institute, which had previously assisted in evaluating models for safety risks.

The stark reminder that the US government could unilaterally revoke access to frontier-grade AI has created an opening for Mistral, a Europe-based lab with open-weight models.

“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.

Reluctant to be pigeonholed into serving only its domestic market, Mistral is eager to emphasize that a lack of fine-grained control over access to AI models could be a problem wherever a business is located—even in the US. By relying on a proprietary model to help repel cyberthreats, Lample says, a business risks the sudden collapse of its defenses.

“Sometimes, people like to [make a big deal] over the US, versus Europe, versus China. But what really matters is to own the model—even for US companies,” says Lample. “If you use a closed model, there is no guarantee it will still be there tomorrow.”