Most enterprise AI projects fail not because companies don’t have the technology, but because they don’t understand how the models they use work. Models are often taught on the web rather than decades of internal documentation, workflows, and institutional knowledge.
Where is that gap? MistralA French AI startup sees an opportunity. On Tuesday, the company announced the Mistral Forge platform, which allows enterprises to build custom models trained on their own data. Mistral announced the platform Nvidia GTCNvidia’s annual technology conference focused more on artificial intelligence and agent models for the enterprise this year.
It’s a dramatic move for Mistral, which has built its business on enterprise customers while its rivals OpenAI and Anthropic have advanced in terms of consumer adoption. CEO Arthur Mensch says Mistral’s laser focus on the enterprise is working: The company is in that direction exceeded $1 billion in annual recurring revenue this year.
Mistral says a big part of doubling down on the enterprise is giving companies more control over their data and AI systems.
“What Forge does is allow enterprises and governments to customize their AI models for their specific needs,” Elisa Salamanca, head of product at Mistral, told TechCrunch.
Several companies in the enterprise AI space already claim to offer similar capabilities, but most focus on fine-tuning existing models or overlaying proprietary data through methods such as advanced generation search.RAG). These approaches do not fundamentally retrain the models; instead, they match or query them at runtime using company data.
Mistral, on the other hand, says it allows companies to build a model from scratch. In theory, this could address some of the limitations of more common approaches—for example, better handling of non-English or highly domain-specific data and more control over model behavior. It can also enable companies to train agent systems using reinforcement learning and reduce reliance on third-party model providers, avoiding risks such as model change or obsolescence.
Techcrunch event
San Francisco, CA
|
October 13-15, 2026
Forge customers can build their own custom models using Mistral’s extensive library of open-weight AI models, which includes the recently introduced smaller models. Mistral Small 4. According to Timothée Lacroix, co-founder and chief technology officer of Mistral, Forge can help you get more value out of their existing models.
“The trade-off we make when building small models is that they can’t be as good at every subject as their larger counterparts, and so the ability to customize them allows us to choose what we emphasize and what we leave out,” Lacroix said.
Mistral recommends which models and infrastructure to use, but both decisions rest with the customer, Lacroix said. For teams that need more than management, Forge comes in Mistral’s forward-thinking engineering team engaging directly with customers to uncover the right information and adapt to their needs – a model borrowed from companies like IBM and Palantir.
“As a product, Forge already comes with all the tools and infrastructure so you can build synthetic data pipelines,” Salamanca said. “But understanding how to build law evaluates and making sure you have the right amount of data is something businesses typically lack the expertise to do, and that’s what FDEs bring to the table.
Mistral has already introduced Forge to partners such as Ericsson, the European Space Agency, Italian consultancy Reply and Singapore’s DSO and HTX. Early adopters include Dutch chip maker ASML Mistral C series A round worth 11.7 billion euros (about $13.8 billion at the time) last September.
These collaborations are emblematic of what Mistral expects Forge’s primary use cases to be. According to Marjorie Janiewicz, Mistral’s chief revenue officer, these include governments needing to develop models that suit their languages and cultures; financial players with high compliance requirements; manufacturers with customization needs; and technology companies that need to adapt the models to the code base.
