Mistral pushes ‘build-your-own AI’ strategy to compete in the enterprise market

Many enterprise AI initiatives fall short not because of a lack of technology, but because the models used fail to understand the businesses they are meant to serve truly. These models are typically trained on general internet data rather than the extensive internal knowledge companies possess, such as documents, workflows, and years of institutional expertise.

This gap presents an opportunity that Mistral, the French AI startup, is aiming to address. On Tuesday, the company introduced Mistral Forge, a platform that enables enterprises to create custom AI models trained on their proprietary data. The announcement was made at Nvidia GTC, Nvidia’s annual technology conference, which this year focuses strongly on artificial intelligence and agent-driven enterprise solutions.

The move highlights Mistral’s strategy of concentrating on enterprise clients, even as competitors like OpenAI and Anthropic have gained significant traction in the consumer space. CEO Arthur Mensch stated that this enterprise-focused approach is yielding results, with the company on track to exceed $1 billion in annual recurring revenue this year.

A central element of this strategy is giving organisations greater control over both their data and AI systems.

“What Forge does is it lets enterprises and governments customise AI models for their specific needs,” said Elisa Salamanca, Mistral’s head of product.

While other companies in the enterprise AI sector claim to offer similar capabilities, most rely on methods such as fine-tuning existing models or applying techniques like retrieval augmented generation (RAG) to incorporate proprietary data. These approaches adapt models or query them dynamically but do not fundamentally retrain them.

Mistral, however, says its platform enables organisations to train models from the ground up. This approach could help overcome some of the limitations of existing methods, such as improving performance with non-English languages or highly specialised domains, while also giving companies more control over how models behave. It may also support the development of agent-based systems using reinforcement learning and reduce dependence on third-party providers, mitigating risks associated with model updates or discontinuation.

With Forge, customers can build custom models using Mistral’s collection of open-weight AI models, which includes smaller options like the recently released Mistral Small 4. According to co-founder and chief technologist Timothée Lacroix, the platform allows companies to extract more value from these models.

“The trade-offs that we make when we build smaller models are that they just cannot be as good on every topic as their larger counterparts, and so the ability to customise them lets us pick what we emphasise and what we drop,” Lacroix explained.

Mistral provides guidance on selecting models and infrastructure, but the final decisions remain with the customer, Lacroix noted. For organisations that need more hands-on support, Forge includes a team of forward-deployed engineers who work directly with clients to identify the right data and tailor solutions to their needs. This approach is similar to strategies used by companies like IBM and Palantir.

“As a product, Forge already comes with all the tooling and infrastructure so you can generate synthetic data pipelines,” Salamanca said. “But understanding how to build the right evals and making sure that you have the right amount of data is something that enterprises usually don’t have the right expertise for, and that’s what the FDEs bring to the table.”

Mistral has already rolled out Forge to several partners, including Ericsson, the European Space Agency, the Italian consulting firm Reply, and Singapore’s DSO and HTX. Early adopters also include ASML, the Dutch semiconductor company that led Mistral’s Series C funding round last September, valuing the company at €11.7 billion (around $13.8 billion at the time).

These collaborations reflect the primary use cases Mistral envisions for Forge. According to chief revenue officer Marjorie Janiewicz, these include government organisations that need models tailored to specific languages and cultural contexts, financial institutions with strict compliance requirements, manufacturers requiring high levels of customisation, and technology companies seeking to align AI models with their codebases.