French artificial intelligence startup Mistral has announced Mistral Forge, a new platform that enables enterprises to build custom AI models trained on their own data. The company unveiled the platform at Nvidia GTC, the chipmaker’s annual technology conference, which this year has placed a strong emphasis on AI and agentic models for enterprise use.
According to Mistral, Forge is designed to bridge the gap between generic AI systems and enterprise-specific requirements. Instead of relying on broad, publicly available datasets, the platform allows organizations to train models that understand their internal context—embedded within systems, workflows, and policies—thereby aligning AI more closely with their unique operations.
The launch also reflects Mistral’s broader strategic positioning. While competitors such as OpenAI and Anthropic have surged ahead in consumer adoption, Mistral has focused heavily on corporate clients. CEO Arthur Mensch said this enterprise-first approach is paying off, with the company on track to surpass $1 billion in annual recurring revenue this year.
READ: Meta delays ‘Avocado’ AI model launch to May: Report (
Mistral says that a big part of doubling down on enterprise is giving companies more control over their data and their AI systems. “What Forge does is it lets enterprises and governments customize AI models for their specific needs,” Elisa Salamanca, Mistral’s head of product, told TechCrunch.
Although several companies in the enterprise AI space already offer similar capabilities, many rely on fine-tuning existing models or layering proprietary data through techniques such as retrieval augmented generation (RAG). This approach typically adapts models at runtime using company data rather than fundamentally retraining them.
Mistral, by contrast, says Forge enables organizations to train models from scratch. In theory, this could address some of the limitations associated with more common approaches, including improved handling of non-English or highly domain-specific data, as well as greater control over model behavior. It may also allow companies to train agentic systems using reinforcement learning while reducing reliance on third-party model providers, thereby avoiding risks such as model changes or deprecation.
Forge customers can build their custom models using Mistral’s extensive library of open-weight AI models, which includes smaller models such as the recently introduced Mistral Small 4. Mistral co-founder and chief technologist Timothée Lacroix said the platform can help unlock additional value from these existing models.
READ: Inside Moltbook: The social platform Meta just acquired (
“The trade-offs that we make when we build smaller models is that they just cannot be as good on every topic as their larger counterparts, and so the ability to customize them lets us pick what we emphasize and what we drop,” Lacroix said.
Lacroix added that while Mistral provides guidance on selecting models and infrastructure, the final decisions remain with the customer. For organizations that require more hands-on support, Forge also includes Mistral’s team of forward-deployed engineers, who work directly with clients to identify relevant data and tailor systems to their needs. This approach draws on a model used by companies such as IBM and Palantir.
Mistral has already rolled out Forge to several organizations, including Ericsson, the European Space Agency, Italian consulting firm Reply, and Singapore’s DSO and HTX, signaling early enterprise adoption of the platform.