Mistral makes its open model case with Mistral Large 4

Mistral, the EU's flagship model provider, is making its case for open weight model leadership and AI sovereignty with the launch of Mistral Large 4, or "Le chonk."

The company launched Mistral Large 4 (ML4) in public preview and made the case in its blog that the model can outperform popular open weight models from the likes of Z.ai, DeepSeek and Moonshot AI. The US open model ecosystem is lacking but players like Nvidia with its Nemotron franchise and Meta with its Muse lineup are looking to close the gap with China.

Mistral wants to be seen as an open model option too and should be given that its input pricing for 1 million tokens is $1.36 and output per million is $4.18. In addition, ML4 is a 1 trillion -parameter multimodal model with 49 billion active parameters.

Simply put, the open model race is often framed as a US vs. China battle, but the reality for enterprises is it’s more western economy models vs. China.

In addition, Mistral is pushing the frontier labs and noted that ML4 was trained from scratch on 3,800 Nvidia Grace Blackwell GPUs in its own data centers in Europe. The company said (emphasis mine):

"The model demonstrates exceptional performance across coding, agentic workflows, and multimodal understanding. It already achieves performance competitive with the strongest open-source models globally, while significantly outperforming any open-weight model developed in the US or Europe. On critical enterprise workloads, including cybersecurity, finance and law, we find it to be state-of-the-art among open models. In some domains such as visual grounding, it goes further still, surpassing even frontier closed models."

• Mistral said it raised €3 billion in Series D funding
• Nvidia’s Nemotron hit enterprise AI inflection point
• Why you should arbitrage your AI models like your vendors do
• Nvidia’s $12.93 billion Hugging Face acquisition validates open weight models

Note the enterprise spin here--especially given many enterprises are looking to run open models and train them with proprietary data. Mistral's ML4 will be seen as a safe choice in Europe and US just like Nvidia Nemotron. Should Mistral get SaaS vendors on board, ML4 will gain share. Mistral's ML4 will also be seen as a good open model play for multinational enterprise given it was trained on more than 160 languages.

According to Mistral, ML4 weights will be released by the end of the month. "Until then, we are red-teaming the model in real-world settings with cybersecurity leaders, vetted partners, and state authorities, who will access the same model with reduced moderation and expanded cyber capabilities," said Mistral.

ML4 is competitive on multiple fronts including cybersecurity and coding, but on automation the model outperforms China's best offerings. On knowledge work, ML4 can top OpenAI's GPT-6 Astra.

The biggest takeaway here is that Mistral took its €3 billion Series D funding and hit the ground running on ML4 and its compute capacity. Now Mistral has to build out the ML4 family.

• Open-Weight Models Are Gaining Ground in Enterprise AI
• Get ready for US open LLMs and just in time
• Rightsizing open models may cut your AI inference spend
• The most important LLM benchmark today: Price