Mistral AI has launched Mistral Forge, a new platform for enterprises that want more control over how AI models are trained, adapted, and deployed. The company is pitching Forge as a way for organizations to move beyond generic AI systems and build models around their own data, internal policies, and infrastructure requirements for an improved enterprise AI adoption strategy.
Many enterprise AI deployments still rely on general-purpose models customized through fine-tuning or retrieval-augmented generation. Forge pushes further than that.
According to VentureBeat’s report on the launch, the platform supports the full training lifecycle, including pre-training on internal datasets, supervised fine-tuning, preference optimization, and reinforcement learning for ongoing improvement.
Forge goes beyond basic fine-tuning
Forge gives enterprises a broader set of tools for building custom models than the standard fine-tuning route. Mistral says its services range from full pre-training with custom data mixtures to continued pre-training, supervised fine-tuning, and alignment work tailored to specific business needs.
On its custom model training page, the company says customers can develop models for specialized use cases in finance, healthcare, manufacturing, and the public sector, and deploy them in environments they control.
Mistral is also emphasizing sovereignty and privacy. The company says customers can run models on-prem, in private or public clouds, or on-device, which is likely to appeal to organizations that have been cautious about sending sensitive data into outside systems. Its positioning is less about convenience and more about control: control over data, control over deployment, and control over how a model behaves in production.
Adoption may be selective, but the pitch is clear
Forge still looks like a platform best suited to enterprises with the budget, data maturity, and technical resources to support custom training. Computerworld reported that analysts expect near-term adoption to be limited, even as the product gives enterprises a more flexible alternative to off-the-shelf AI systems. Full-cycle model training is a heavier lift than plugging into an API, and many companies will still prefer faster, simpler options.
Even so, Forge adds a notable option to the enterprise AI market. Mistral is betting that some organizations want more than prompt engineering layered on top of a third-party model. They want systems trained on proprietary knowledge, aligned with internal policies, and deployed in environments they govern themselves. For companies in tightly regulated sectors, or for those trying to protect domain expertise and sensitive data, that pitch is likely to resonate.
Also read: Nvidia’s GTC 2026 keynote underscores how quickly the infrastructure behind enterprise AI is evolving.