Microsoft has made OpenAI’s new free and open GPT model, gpt-oss-20b, available to Windows 11 users through Windows AI Foundry, its platform that allows local AI feature and API use on PCs. The lightweight gpt-oss-20b model, optimized for tasks like code execution and tool use, requires a PC with at least 16GB of VRAM, supported by modern Nvidia or Radeon GPUs. Microsoft said the model is ideal for building autonomous assistants and embedding AI into workflows, even in environments with limited bandwidth.
This release marks the first time an OpenAI model can run locally on Windows computers, providing developers with greater flexibility for AI deployment on edge devices. Microsoft plans to expand support to more devices, including macOS, soon. Furthermore, both gpt-oss-20b and the larger gpt-oss-120b model are also available on Microsoft’s Azure AI Foundry cloud platform, enabling secure and efficient AI workloads across cloud and edge.
OpenAI’s gpt-oss-20b is a text-only model trained with reinforcement learning for tool use, but it can have higher rates of hallucination compared to commercial models. Still, its availability under an open-weight Apache 2.0 license empowers developers to run, fine-tune, and customize the model locally or cloud-based with fewer restrictions. Amazon’s AWS has also made these models available, illustrating the growing accessibility of OpenAI’s open-weight technology.
This development fits into Microsoft’s broader strategy of integrating advanced AI capabilities locally on Windows, complementing its previous AI initiatives and furthering its competitive edge in AI infrastructure.