Mistral Forge: How the French AI Startup Is Empowering Enterprises to Build Custom AI Models from the Ground Up

Mistral Takes Aim at Enterprise AI With a Bold New Strategy

Most enterprise AI initiatives stumble — not because companies lack access to cutting-edge technology, but because the underlying models simply don't speak the language of their business. Generic models trained on publicly available internet data often miss the nuance buried in decades of internal documentation, operational workflows, and hard-earned institutional knowledge.

That critical disconnect is exactly where French AI startup Mistral is planting its flag.

Introducing Mistral Forge

At Nvidia's annual GTC technology conference — an event this year centered heavily on AI and agentic enterprise systems — Mistral unveiled Mistral Forge, a first-of-its-kind platform designed to give organizations the power to build and train AI models entirely on their own proprietary data.

Unlike most competing solutions that rely on fine-tuning pre-existing models or layering company data on top through techniques like Retrieval Augmented Generation (RAG), Mistral Forge takes a fundamentally different approach: training models from scratch.

"What Forge does is it lets enterprises and governments customize AI models for their specific needs," said Elisa Salamanca, Mistral's Head of Product.

Why Training From Scratch Matters

The distinction between fine-tuning and ground-up training may sound technical, but its business implications are significant. Fine-tuning and RAG-based methods adapt or query existing models at runtime using company data — they don't fundamentally reshape how a model thinks or processes information.

By enabling full model training from the ground up, Mistral Forge offers several potential advantages:

• Better handling of non-English and highly specialized data
• Greater control over model behavior and outputs
• The ability to train agentic systems using reinforcement learning
• Reduced dependence on third-party model providers, eliminating risks associated with unexpected model changes or deprecation

Forge customers can build their custom models by drawing from Mistral's extensive library of open-weight AI models, which includes compact options such as the newly launched Mistral Small 4.

Smaller Models, Smarter Customization

According to Mistral co-founder and Chief Technologist Timothée Lacroix, Forge unlocks new value from the company's existing model lineup by allowing businesses to tailor what each model prioritizes.

"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 explained.

While Mistral provides guidance on which models and infrastructure best suit a customer's needs, both decisions ultimately remain in the hands of the client.

Forward-Deployed Engineers: A Palantir-Inspired Model

For companies that need more than technical tooling, Mistral Forge comes bundled with a team of forward-deployed engineers (FDEs) who work directly alongside customers. This hands-on service model draws inspiration from enterprise veterans like IBM and Palantir.

These embedded engineers help organizations surface the right data, build appropriate evaluation frameworks, and ensure training pipelines are properly configured — expertise that most enterprises simply don't have in-house.

"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," Salamanca noted.

The platform itself also comes equipped with built-in tooling for generating synthetic data pipelines, further lowering the barrier to entry for organizations embarking on custom model development.

Early Partners and Target Industries

Mistral has already onboarded a notable roster of early Forge partners, including:

• Ericsson
• The European Space Agency
• Reply (Italian consulting firm)
• DSO and HTX (Singapore-based organizations)
• ASML — the Dutch semiconductor giant that led Mistral's Series C funding round last September, valuing the company at approximately €11.7 billion (around $13.8 billion at the time)

According to Chief Revenue Officer Marjorie Janiewicz, Forge is being targeted at four primary sectors:

1. Governments requiring models adapted to specific languages and cultural contexts
2. Financial institutions operating under strict compliance requirements
3. Manufacturers with deep product and process customization needs
4. Technology companies seeking models finely tuned to their proprietary codebases

Enterprise Focus Fueling Rapid Growth

While rivals OpenAI and Anthropic have captured mainstream consumer attention, Mistral CEO Arthur Mensch has remained unwavering in the company's enterprise-first approach — and the numbers suggest it's paying off. Mistral is reportedly on track to surpass $1 billion in annual recurring revenue this year.

The launch of Mistral Forge represents a doubling down on that strategy, offering corporate clients and governments not just access to powerful AI, but genuine ownership and control over the systems they deploy.