Mistral targets enterprise AI with new custom model platform

On Tuesday, French AI startup Mistral announced Mistral Forge at the Nvidia GTC technology conference. The platform is designed to let enterprises and governments customize AI models for their specific needs by training them from scratch on their own data. This approach contrasts with more common methods that rely on standard fine-tuning or Retrieval Augmented Generation (RAG)—a technique used to layer proprietary data on top of existing models. According to the company, most enterprise AI projects fail because the models they are using do not understand their business, as they are typically trained on general internet data rather than internal workflows and institutional knowledge.

To address this, Forge allows customers to build custom models using Mistral’s library of open-weight AI models—where the model weights are publicly available—which includes smaller models like Mistral Small 4. According to co-founder and chief technologist Timothée Lacroix, Forge can help unlock more value out of its existing models. “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.

The platform also includes tooling for synthetic data and provides access to forward-deployed engineers (FDEs)—engineers who embed directly with customers to help surface data and adapt models. This hands-on business model mirrors strategies used by companies like IBM and Palantir. Elisa Salamanca, Mistral’s head of product, noted that while Forge comes with the tooling and infrastructure to generate synthetic data pipelines, FDEs bring the necessary expertise to help enterprises build the right evaluations and manage their data.

The launch comes as Mistral experiences significant commercial growth. CEO Arthur Mensch stated that the company is on track to surpass $1 billion in annual recurring revenue this year. This growth is backed by major European partners; last September, Dutch chipmaker ASML led Mistral’s Series C funding round, which valued the startup at €11.7 billion (approximately $13.8 billion at the time). Mistral has already made Forge available to early adopters and partners, including Ericsson, the European Space Agency, Italian consulting company Reply, and Singapore’s DSO and HTX.

Why it matters

Mistral Forge attempts to solve the “last mile” problem in enterprise AI—where generic models fail to grasp specific business contexts—by moving beyond simple Retrieval Augmented Generation to full custom model training.