Mistral Forge is an enterprise AI platform launched by Mistral AI at NVIDIA GTC 2026 on March 17, 2026. It enables organizations to build frontier-grade custom AI models trained exclusively on their proprietary data, going far beyond simple fine-tuning. The platform supports the full model lifecycle including pre-training on large internal datasets, post-training via supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF), direct preference optimization (DPO), and low-rank adaptation (LoRA) for efficient specialization. Forge supports both dense and Mixture-of-Experts (MoE) architectures, enabling organizations to optimize for performance, cost, and latency. It also handles multimodal inputs including text and images. A standout capability is its data pipeline tooling for data acquisition, curation, and synthetic data generation. The platform features agent-first design where autonomous agents can fine-tune models, optimize hyperparameters, schedule training jobs, and monitor evaluation metrics. Deployment is flexible across private cloud, on-premises infrastructure, or Mistral compute, with strict data isolation and compliance-aligned governance. Forge is compatible with OpenAI's API schema for easy integration. Enterprise partners already include ASML, Ericsson, European Space Agency, DSO National Laboratories Singapore, and Reply.