Microsoft brings OpenAI’s new open model gpt-oss-20b to Windows AI Foundry for local AI power

OpenAI has released a new free and open-source GPT model capable of running directly on personal computers, and Microsoft has quickly made it accessible to Windows users through the Windows AI Foundry. The lightweight gpt-oss-20b model is now available on Windows, with support for macOS expected soon. To run the model locally, users will need a PC or laptop equipped with at least 16GB of VRAM, meaning systems with high-end GPUs from Nvidia or select Radeon models are required. Optimized for code execution and tool use, the gpt-oss-20b is designed to power autonomous assistants and integrate AI into real-world workflows—especially in environments with limited internet bandwidth. Microsoft has pre-optimized the model for local inference on Windows, streamlining the setup process. The company also hinted that additional device support is on the way, potentially including a more tailored version for Copilot Plus PCs in the future, continuing its recent trend of embedding local AI capabilities into Windows. This rapid integration marks the first time OpenAI’s latest model is available for local use on Windows, but it also reflects a broader shift in the tech landscape. Amazon has similarly moved quickly to incorporate the new open-weight GPT-OSS models into its cloud infrastructure, giving its platform direct access to the latest OpenAI advancements. This development adds a new layer to the complex relationship between Microsoft and OpenAI, as well as the competitive dynamics between Microsoft and Amazon in the AI space. While Microsoft remains a key partner and investor in OpenAI, the availability of open models on rival cloud platforms signals a growing trend toward wider access and decentralization of AI capabilities.