NVIDIA Nemotron

NVIDIA Nemotron is a family of highly efficient, multimodal, open AI models built for long-running, self-evolving agents, designed for fast task completion with high reasoning throughput and leading accuracy for complex agent workflows. NVIDIA publishes the training datasets, techniques and model weights openly, and the family includes models for reasoning (Nano, Super and Ultra sizes), visual understanding, speech, retrieval-augmented generation (RAG), and safety, available as optimized NVIDIA NIM microservices for enterprise deployment.