Microsoft makes OpenAI’s latest GPT-OSS model available on Windows

OpenAI’s new open GPT model is now accessible to Windows users, giving developers and AI enthusiasts the ability to run advanced AI locally without relying solely on cloud services. The lightweight gpt-oss-20b model can run directly on PCs equipped with powerful GPUs, making it easier to create autonomous assistants, integrate AI into workflows, and experiment with local AI solutions. This marks the first time Windows users can natively run an OpenAI model on their own devices.

How the gpt-oss-20b model works on Windows

The gpt-oss-20b model is designed for local AI inference, meaning it can run without a continuous internet connection. Microsoft recommends a PC or laptop with at least 16GB of VRAM, compatible with high-end Nvidia or AMD Radeon GPUs. Optimized for code execution and real-time tool usage, this model empowers users to develop AI-powered tools that work efficiently even in bandwidth-limited environments.

Why Microsoft’s local AI move matters

By bringing OpenAI’s latest open model to Windows AI Foundry, Microsoft is making local AI development more accessible than ever. This move supports developers looking to experiment with autonomous applications while reducing cloud dependency. It also aligns with Microsoft’s broader AI strategy of integrating more on-device models into Windows, following the trend of Copilot Plus PCs and other local AI initiatives.

What’s next for AI on Windows

Microsoft plans to expand device compatibility and optimize gpt-oss-20b for a wider range of hardware in future updates. With macOS support on the horizon, and major cloud competitors like Amazon also adopting the GPT-OSS models, this release signals a new phase in local AI computing. For Windows users, it opens the door to experimenting with AI applications faster, more securely, and with more control over resources.