How to Use OpenAI's Latest AI Models on NVIDIA RTX Graphics Cards

OpenAI and NVIDIA have worked together to make OpenAI’s newest open source artificial intelligence models run faster and smarter on NVIDIA GeForce RTX and RTX PRO graphics cards. These models are designed to help with tasks like web searching and research. Now, you can use them both on powerful cloud computers and on your own PC.

The new models, called gpt oss 20b and gpt oss 120b, are now available to everyone. AI fans and developers can try them out on their NVIDIA RTX desktop computers or workstations using simple tools such as Ollama, llama cpp, and Microsoft AI Foundry Local. For example, you can reach impressive speeds with the latest NVIDIA GeForce RTX 5090.

These models use advanced technology that allows for flexible thinking and reasoning. They follow instructions well, can use tools, and handle longer text inputs than many other AI models. They have been trained using powerful NVIDIA H100 chips and can manage up to one hundred thirty one thousand context lengths, making them great for things like reading documents, helping with coding, and diving deep into research.

NVIDIA’s RTX graphics cards are the first to support these models with a technology called MXFP4. This makes your computer use its resources efficiently while still giving you excellent AI performance.

Getting started is easy. The Ollama app provides a simple way to chat with these new AI models on your RTX graphics card. You just install the app, pick a model from a menu, and start sending messages. Ollama also lets you attach image or text files and adjust how much information the model remembers during a conversation. Developers can use special tools in Ollama to add these AIs into other software.

You are not limited to Ollama. You can also use the gpt oss models through programs like llama cpp and libraries designed for RTX, or with Microsoft AI Foundry Local on Windows. NVIDIA continues to work with open source communities to make sure these models work well and run fast on RTX graphics cards.

Thanks to these efforts, more people can now try out and build advanced AI solutions on their own Windows PCs and workstations. If you want to learn more or get support, you can visit NVIDIA's blogs, join their Discord community, or follow them on social media for tips and updates.

Original article and image: https://blogs.nvidia.com/blog/rtx-ai-garage-openai-oss/