The AI startup Mistral introduced Mistral Forge, a platform designed to help enterprises build AI models trained on their own data. The announcement was made at Nvidia GTC, the chipmaker’s annual conference, which this year highlights enterprise AI and agentic systems.
The platform was developed because many enterprise AI initiatives struggle despite the availability of technology, as the models fail to reflect the specific needs of the businesses using them. Most systems are trained on broad internet data rather than internal company knowledge, processes, and documentation.
The launch underscores Mistral’s enterprise-focused strategy, even as competitors like OpenAI and Anthropic lead in consumer markets. CEO Arthur Mensch said the approach is paying off, with the company expecting to exceed $1 billion in annual recurring revenue this year.
Mistral says Forge gives organizations greater control over both their data and 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.
While other vendors offer similar tools, many rely on methods like fine-tuning or retrieval augmented generation (RAG), which adapt existing models without fully retraining them. Mistral claims its approach goes further by enabling companies to build models from the ground up.
This could improve performance on specialized or non-English data and give businesses more control over model behavior. It may also support the development of agentic systems using reinforcement learning while reducing reliance on external model providers.
Forge allows customers to use Mistral’s library of open-weight models, including smaller systems such as Mistral Small 4. Customization can be especially helpful in overcoming the limitations of smaller models, according to co-founder and chief technologist Timothée Lacroix.
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.
Mistral provides guidance on model and infrastructure choices, though final decisions remain with the client. The platform offers support from forward-deployed engineers who work directly with customers to tailor solutions — an approach similar to 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.
Forge is already being used by partners such as Ericsson, the European Space Agency, Reply, and Singapore’s DSO and HTX. Early adopters also include ASML, which led Mistral’s Series C round last September at a €11.7 billion valuation.
According to chief revenue officer Marjorie Janiewicz, the platform is expected to be especially useful for governments needing localized AI, financial institutions with strict compliance demands, manufacturers requiring customization, and tech firms adapting models to their codebases.
Image: Elisa Salamanca LinkedIn profile; Timothee Lacroix LinkedIn profile; Arthur Mensch LinkedIn profile/ Edited 18.03.2026
