Most enterprise AI projects fail due to a lack of understanding of the business by the models, not because of technology shortages. Models are often trained on internet data instead of internal resources.
That’s where Mistral, a French AI startup, sees potential. The company announced Mistral Forge, a platform for enterprises to build custom models using their own data. Mistral announced this platform at Nvidia GTC, a tech conference focused on AI and agentic models.
Mistral focuses on corporate clients, unlike other companies like OpenAI and Anthropic, which focus on consumer adoption. CEO Arthur Mensch says Mistral is on track to exceed $1 billion in annual recurring revenue this year.
Mistral emphasizes giving companies more control over their data and AI systems. “Forge lets enterprises and governments customize AI models for their specific needs,” said Mistral’s head of product, Elisa Salamanca.
While other companies offer similar capabilities, they usually refine existing models by adding proprietary data. Mistral, however, enables companies to train models from scratch, offering better handling of non-English or domain-specific data and reducing reliance on third-party providers.
Forge clients can create custom models using Mistral’s library of open-weight AI models. Co-founder Timothée Lacroix said it helps maximize existing models’ value by allowing for customization.
Customers make decisions on models and infrastructure with advice from Mistral, while forward-deployed engineers assist in adapting to customer needs. Forge includes tools for synthetic data pipelines, which helps enterprises build necessary infrastructure.
Forge has been made available to partners like Ericsson, the European Space Agency, and others. Early adopters include ASML, a Dutch chipmaker. Forge’s use cases include governments customizing models for local languages, financial institutions with compliance needs, manufacturers needing customization, and tech companies tuning models to their code.