Nvidia, which is the main supplier of powerful chips for third-party AI developments, has taken a step towards becoming a serious player in the AI modeling industry itself.
Nvidia, which is the main supplier of powerful chips for third-party AI developments, has taken a step towards becoming a serious player in the AI modeling industry itself.
As Wired reports, Nvidia is releasing a series of advanced open-source Nvidia Nemotron 3 models, and is also providing engineers with data and tools to help them use them.
«Open innovation is the foundation of AI progress. With Nemotron, we are transforming cutting-edge AI into an open platform that gives developers the transparency and efficiency they need to build agent-based systems at scale,» said Nvidia CEO Jensen Huang.
Nvidia’s new Nemotron 3 models are among the most powerful, according to benchmark results published by the company ahead of release. They can be downloaded, modified, and run on your own hardware.
«The move comes as AI companies like OpenAI, Google, and Anthropic are developing their own increasingly powerful chips. It could act as a buffer if those firms eventually abandon Nvidia’s technology,» Wired reports.
Nvidia is taking a more transparent approach than many of its American competitors, publishing the data used to train Nemotron, which should make it easier for third-party engineers to modify the models.
Nemotron 3 models come in three versions: Nano, which has 30 billion parameters; Super, which has 100 billion parameters; and Ultra, which has 500 billion parameters.
«We believe open source is the foundation for AI innovation that continues to accelerate the global economy,» said Kari Anne Briskey, vice president of generative AI software at Nvidia.
As the publication notes, over the past year, American AI companies have moved away from openness, becoming less willing to tell competitors about their latest engineering tricks, while Chinese companies, including DeepSeek, Alibaba, Moonshot AI, Z.ai and MiniMax, regularly release powerful open models and publish detailed information about them, making their offerings more attractive to third-party engineers.
Previously, dev.ua wrote that Nvidia’s AI model outperformed the top OpenAI and Anthropic models in tests .
Meanwhile, it has also become known that Nvidia is implementing a chip tracking system to combat smuggling.
