NVIDIA has launched Nemotron 3 Super, a 120‑billion‑parameter open model with 12 billion active parameters designed to power complex agentic AI systems at scale. The model, available immediately, integrates advanced reasoning capabilities to complete tasks efficiently and accurately for autonomous agents. It features a 1‑million‑token context window to maintain full workflow memory and prevent goal drift in multi‑agent applications.
Nemotron 3 Super employs a hybrid mixture‑of‑experts architecture that delivers up to five times higher throughput and twice the accuracy of its predecessor. Running in NVFP4 precision on the NVIDIA Blackwell platform, it achieves up to four times faster inference than FP8 on NVIDIA Hopper without accuracy loss. The model has achieved top rankings on Artificial Analysis and DeepResearch Bench leaderboards for efficiency and reasoning coherence. NVIDIA is releasing it with open weights under a permissive license, enabling deployment across workstations, data centers, and cloud environments.
Industry partners including Perplexity, Amdocs, Palantir, Siemens, and Dell Technologies are integrating Nemotron 3 Super into their AI ecosystems to enhance search, software development, cybersecurity, and manufacturing automation. The model is accessible through build.nvidia.com, OpenRouter, and Hugging Face.
Launched today, NVIDIA Nemotron 3 Super is a 120‑billion‑parameter open model with 12 billion active parameters designed to run complex agentic AI systems at scale. Available now, the model combines advanced reasoning capabilities to efficiently com
The ‘1 Nojor’ media platform is now live in beta, inviting users to explore and provide feedback as we continue to refine the experience.
