In a significant strategic shift, the American chipmaking titan Nvidia has unveiled its latest family of open-source artificial intelligence models, dubbed Nemotron 3. Announced on December 19, 2025, this launch marks a deeper commitment to transparent AI development, setting the company on a potential collision course with its own largest customers.
Nemotron 3: Three Models for Diverse AI Tasks
The newly released Nemotron 3 suite is described by Nvidia as its most efficient family of open models to date. It is specifically engineered for building and deploying precise AI agents. The family comprises three distinct models, each tailored for different levels of complexity.
The lineup includes the Nemotron 3 Nano, a 30-billion parameter model optimized for targeted tasks. For more demanding multi-agent applications, there is the Nemotron 3 Super with 100 billion parameters. At the top end sits the Nemotron 3 Ultra, a massive 500-billion parameter model designed to handle the most complex AI challenges.
According to benchmark data shared by Nvidia, the Nemotron 3 Nano demonstrates superior accuracy compared to Alibaba's similarly sized Qwen model. It also reportedly delivers better inference throughput than OpenAI's GPT-OSS-20B model.
Transparency as a Strategy in a Competitive Landscape
The most notable aspect of this release is Nvidia's unprecedented transparency. The company has publicly released the complete training dataset, model weights, training recipe, and the underlying framework for Nemotron 3. This open-source approach starkly contrasts with the more guarded strategies of several US-based rivals.
"Open innovation is the foundation of AI progress," stated Nvidia CEO Jensen Huang. "With Nemotron, we're transforming advanced AI into an open platform that gives developers the transparency and efficiency they need to build agentic systems at scale."
This move is particularly intriguing as it unfolds against a backdrop where Nvidia's primary customers—tech giants like OpenAI, Google, Meta, and Anthropic—are actively pursuing in-house chip development to reduce their long-term dependence on Nvidia's GPUs. By positioning itself as a leading open-source AI model maker, Nvidia is diversifying its influence in the AI ecosystem beyond just hardware.
Tools, Acquisitions, and the Open-Source Ethos
Complementing the Nemotron 3 models, Nvidia has released a suite of tools to assist developers in customizing and fine-tuning AI models. The company also launched specialized libraries to train AI agents for autonomous task handling using techniques like reinforcement learning.
Further cementing its commitment to the open-source high-performance computing community, Nvidia announced the acquisition of SchedMD, the creator of the popular open-source workload manager, Slurm. Nvidia has pledged to maintain Slurm as vendor-neutral and open-source software.
"We believe open source is the foundation for AI innovation, continuing to accelerate the global economy," said Kari Ann Briski, Nvidia's vice president of generative AI software for enterprise, in an interview with Wired.
This series of announcements signals Nvidia's ambitious plan to solidify its role not just as the indispensable hardware provider for the AI boom, but as a central architect of the open-source software and models that will power its future.
