Dubbed “Forge,” the system allows enterprises to train models on internal documentation, codebases, structured data and operational records so the models learn each enterprise’s vocabulary, reasoning patterns and constraints, Mistral said in a Tuesday (March 17) press release.
Forge supports pre-training that allows organizations to build domain-aware models, post-training that lets them refine models for specific tasks and environments, and reinforcement learning that helps them improve agentic performance in real environments, according to the release.
By building models trained on their own knowledge, enterprises can retain control over the models, data and intellectual property; build agents that can navigate internal systems, use tools correctly and make decisions within the organization’s constraints; build both dense models and mixture-of-experts ones; and continuously refine models as needed, the release said.
Mistral expects Forge to support the workflows of government agencies, financial institutions, software teams, manufacturers and large enterprises. For financial institutions, Forge enables models to be trained on compliance frameworks, risk procedures and regulatory documentation so they can produce outputs consistent with each institution’s internal governance policies, per the release.
“AI models are becoming a foundational layer of enterprise infrastructure,” Mistral said in the release. “As organizations integrate AI agents into core operations, the ability to encode institutional knowledge into model behavior will become increasingly important.”
PYMNTS reported in June that Mistral and other AI companies are not only rolling out more advanced iterations of existing models but also unveiling custom models tailored to distinct use cases and specialized needs.
The next round of innovation will come from custom-built intelligence for industries or sectors, the report said.
It was reported in February that Mistral’s base of enterprise customers exceeds 100 companies and includes banking giant HSBC and automaker Stellantis.
Mistral Co-founder and CEO Arthur Mensch said at the time that the company’s annualized revenue run rate was “north of $400 million,” up from $20 million a year earlier.
The company raised 1.7 billion euros (about $2 billion) in a September 2025 Series C funding round that valued it at 11.7 billion euros (about $13.5 billion).
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.