Mistral launches 1-trillion-parameter open weight model

Mistral is launching Mistral Large 4, a 1-trillion-parameter model that the company says is the most powerful open-weight model it has ever built — and the most capable of its type on the market. Nicknamed “Le Chonk,” ML4 is pitched as a frontier tool for two domains in particular: cyber defense and coding, Pierre Stock , Mistral’s vice president of science, said during a press briefing on Oct. 28. He said those capabilities are what companies and governments need most to defend against rising cyber threats. Corentin Petit, the company’s vice president of solutions, said those two concentrations are the use cases that will help customers move the needle. The company says the model’s gains came from reinforcement learning run on its own cluster in France, using just 10 megawatts of power — a fraction of what competitors draw, Stock said. Mistral has raised some 6 billion euros ($6.7 billion) over its three-year history. Stock said rivals have spent hundreds of billions to reach comparable frontier capability. Stock said ML4’s efficiency runs through the whole stack. “Once you’re efficient, you can scale accordingly,” he said, “and you can do similarly or even better than your competitors with less.” [gallery size="full" columns="2" link="none" ids="57390297,57390296"] On Mistral’s coding index, the company said ML4 uses 47% fewer tokens per task on average than GLM 5.3 and that its high-reasoning mode is now 50% more efficient than a preview released three weeks ago. Beyond code and security, the model is tuned for financial-services workloads — processing complex spreadsheets and large volumes of filings for compliance, KYC, fraud, and risk, Petit said. Sovereignty The open-weight release is central to Mistral’s sovereignty pitch that organizations can run ML4 entirely on their own systems, keeping both the model and their data in-house, Stock said. He likened intelligence to electricity — no one outside your organization should be able to switch it off. “You don’t want to live in a world in which your neighbor or a utility can cut off electricity,” he said, adding that once weights are operational, “it’s impossible to cut off access.” The quantized version, which Mistral is also releasing, runs on four B200 or B300 GPUs, Stock said. The appetite is real. In Nvidia’s 2026 State of AI in Financial Services survey of more than 800 professionals, 84% said open-source models and software are important to their AI strategy. Nvidia also reported financial services firms are deploying open-weight models at higher rates than most other sectors. For Petit, that is the opening. He said Mistral can help customers build “a custom model that is better than off-the-shelf frontier on their own tasks at a fraction of the token cost and that nobody can switch off.” Register here for early-bird pricing for the FinAi Banking Summit, taking place March 1-2 in Denver. View the full event agenda here .

Mistral launches 1-trillion-parameter open weight model

Model trained for financial services specific tasks: KYC, fraud detection and documentation

From raw to riches: A lender’s path to data readiness in AI era

Related Posts

5 noteworthy numbers in Anthropic’s IPO filing

AI companies will issue more equity to finance AI buildout

Anthropic flags ‘Existential Risks’ in long-awaited IPO filing

EMERGING FINTECH DIRECTORY

FinAi Podcast

SPONSORED

The Next Generation of Cyberthreats: Powered by AI

Build an Antifragile Strategy to Outperform the Market

How AI and Product Experts Turn Fuzzy Requirements Into Focused Dev-ready Roadmaps

Connect

Welcome Back!

Retrieve your password

Continue Reading with FinAi News Premium - Less than $2/Day

Mistral launches 1-trillion-parameter open weight model

Model trained for financial services specific tasks: KYC, fraud detection and documentation

From raw to riches: A lender’s path to data readiness in AI era

Related Posts

FinAi Podcast

Connect

Welcome Back!

Retrieve your password

Unlock This Article

Continue Reading with FinAi News Premium - Less than $2/Day