Mistral AI unveils Forge enterprise AI training platform

Mistral AI has introduced Forge, a system for enterprises to build AI models grounded in their proprietary knowledge. Most AI models available today are trained primarily on publicly available data and designed to perform well across a broad range of tasks. Enterprises operate using internal knowledge such as engineering standards, compliance policies, codebases, operational processes, and years of institutional decisions. Forge bridges the gap between generic AI and enterprise-specific needs by allowing organizations to train models that understand their internal context. The company has partnered with ASML, DSO National Laboratories Singapore, Ericsson, European Space Agency, Home Team Science and Technology Agency (HTX) Singapore, and Reply. Forge enables enterprises to build models that internalize domain knowledge through training on internal documentation, codebases, structured data, and operational records. The platform supports pre-training, post-training, and reinforcement learning approaches. It offers flexibility with dense and mixture-of-experts architectures as well as multimodal inputs. Forge allows enterprises to retain control over models and data using internal policies and evaluation standards. Custom models enable enterprise agents to interpret internal terminology and follow operational procedures with greater reliability. Mistral AI designed Forge for agents, allowing them to fine-tune models, schedule jobs, and generate synthetic data. The system includes continuous adaptation through reinforcement learning and evaluation against internal benchmarks.