Mistral Releases Trillion-Parameter Model Le Chonk, Claiming Top Open-Weight Performance Outside China
Mistral’s one-trillion-parameter Le Chonk is freely available to use and customize, with a final release expected by the end of October 2026. Trained from scratch, it targets coding, cyberdefense, manufacturing, finance, and electrical engineering.
French AI company Mistral released a new open-weight model in October 2026 that it claims outperforms every comparable model developed outside of China. The model, formally called Mistral Large 4 and nicknamed Le Chonk, contains one trillion parameters and is freely available for anyone to use and customize. A preview version is live now, with a final release expected by the end of the month.
Le Chonk is designed as a general-purpose model but is optimized specifically for coding, cyberdefense, and industry-specific tasks in manufacturing, finance, and electrical engineering. Mistral says it trained the model from scratch — a distinction the company draws against Chinese labs, which the US government has accused of using distillation techniques to close the performance gap with leading proprietary models.
“Mistral is still in the race of getting the best model,” said Guillaume Lample, cofounder and chief scientist at Mistral. “This is the main message.”
Mistral positions Le Chonk as “very, very close” to some proprietary models while being considerably cheaper for businesses to run, since open-weight models cost only as much as the compute they consume. The company argues the release removes the remaining reasons a business might choose a closed proprietary model over an open-source alternative.
The launch comes as Mistral has seen significant momentum. In September, the company raised a $3.3 billion funding round at a $24 billion valuation — reported to be the largest ever raise by a European tech company — and its earnings have reportedly grown roughly 20-fold over the past year. Mistral generates revenue by charging pay-as-you-go fees for cloud-based model access and by deploying engineers to help customers customize models.
The release also arrives amid growing geopolitical tensions over access to frontier-grade AI, with fractures widening between the US and its transatlantic allies on issues including tariff policy and the regulation of American technology firms.