Ted Hisokawa
August 5th, 2025, 17:48
Openai and Nvidia have launched new AI models optimized for Nvidia’s infrastructure, bringing promising advances in AI development and application in a variety of industries.
Openai and Nvidia have announced the release of two new Open-Weight AI Reasoning Models, the GPT-OSS-120B and GPT-OSS-20B. These models are intended to provide access to advanced AI development for developers, businesses and governments around the world. According to Nvidia Newsroom, the collaboration highlights the potential for community-driven innovation and Nvidia’s important role in the democratization of AI technology.
Innovating AI infrastructure
The new model is designed to be used in a variety of fields, including healthcare, manufacturing, and generator AI applications. Trained with Nvidia H100 GPUs, Openai’s model works best on Nvidia’s Cuda platform, supported by millions of GPUs around the world. With these advancements, Nvidia’s Blackwell Architecture offers unprecedented efficiency, achieving 1.5 million tokens per second on the NVIDIA GB200 NVL72 system.
Nvidia Blackwell: Game Changer
Nvidia’s Blackwell Architecture is crucial in supporting the exponential growth of advanced inference models like GPT-Oss. It features innovations such as NVFP4 4-bit accuracy, allowing for high-precision inference with reduced power and memory requirements. This advancement allows for real-time deployment of large-scale language models (LLMs) of Trillion parameters, unlocking economic values that are important to the organization.
Global Access and Development Opportunities
A widely accessible computing infrastructure, Nvidia Cuda will be able to deploy AI models across a variety of platforms, including NVIDIA DGX Cloud and GeForce RTX-powered PCs. With over 450 million CUDA downloads, developers now have access to these latest models optimized for Nvidia’s technology stack. Openai and Nvidia are also partnering with leading open framework providers to provide model optimization and promote diverse development opportunities.
The legacy of innovation
The collaboration is based on the long-standing relationship between Nvidia and Openai, which began with the distribution of the first NVIDIA DGX-1 AI supercomputer in 2016. Since then, the two companies have worked together to advance AI capabilities and have optimized Openai’s GPT-OSS model using NVIDIA’s comprehensive software stack.
By enhancing these models of Nvidia Blackwell and RTX GPUs, Nvidia aims to accelerate AI development for a wider community of developers around the world.
Image source: ShutterStock
