Mistral Claims “Le Chonk” Can Compete with Top AI Models

As the global race for **artificial intelligence** capabilities intensifies, French firm **Mistral** has introduced a new model that it asserts rivals the finest offerings from the **US** and **China**. This innovative model, dubbed **Mistral Large 4**, or **Le Chonk**, boasts an impressive one trillion parameters and is freely accessible for public use and customization.

Currently in its preview phase, with a full launch expected later this month, Le Chonk is specifically optimized for tasks in coding, cyber defense, manufacturing, finance, and electrical engineering. According to **Guillaume Lample**, co-founder and chief scientist at **Mistral**, various specialties remain underexplored by other labs, presenting opportunities for enhancement in model performance.

Mistral claims that Le Chonk stands as the most powerful open-weight model developed outside of China and is “very, very close” in performance to proprietary models. Unlike Chinese labs, which have faced accusations from the US government regarding the unethical training practices of distillation—where smaller models are taught using the outputs of larger models—Mistral asserts that its development process involved training from the ground up.

One of the compelling advantages of open-weight models is their cost efficiency, tied directly to the computational resources they utilize. **Mistral** believes that by narrowing the performance gap with leading proprietary models and offering a viable alternative to Chinese solutions, Le Chonk will alleviate any lingering concerns businesses might have about adopting open-source models.

“Mistral is still in the race to deliver the best model,” Lample emphasized, underlining the company’s commitment to innovation.

Though **Mistral** has historically operated with less funding and fewer computational resources than giants like **OpenAI** and **Anthropic**, it has made significant strides in model performance, revenue, and release frequency. While US firms often demand a premium for their proprietary offerings, **Mistral** generates revenue through a pay-as-you-go system for running models in the cloud and providing engineering support tailored to client needs. Recently, the company gained momentum by raising **$3.3 billion** in a funding round, achieving a valuation of **$24 billion**, marking the largest funding acquisition by a tech company in Europe. Reportedly, Mistral’s revenue has surged by **20 times** over the past year.