Microsoft Integrates OpenAI’s GPT-OSS-20b for AI Functionality on Windows 11: A Lightweight, Text-Based Model with Hallucination Warning

Microsoft has announced the availability of OpenAI’s latest free and open GPT model, named gpt-oss-20b, to Windows 11 users through its AI-focused platform, Windows AI Foundry. This move enables users to access AI features, APIs, and popular open-source models directly on their computers.

In a recent blog post, Microsoft highlighted that the lightweight and tool-savvy gpt-oss-20b is optimized for agentic tasks such as code execution and tool utilization. It operates efficiently across various Windows hardware configurations, with support for additional devices soon. This model is particularly suitable for developing autonomous assistants or integrating AI into real-world workflows, even in bandwidth-restricted environments.

OpenAI’s gpt-oss-20b can run on consumer PCs and laptops equipped with at least 16GB of VRAM, as found in modern GPUs from Nvidia or Radeon. The model has been trained using high-compute reinforcement learning, enabling it to excel at powering AI agents and call tools like web search or Python code execution.

However, the text-only nature of the model means it cannot process or generate images or audio like some of OpenAI’s other models. Furthermore, gpt-oss-20b has been observed to hallucinate in response to 53% of questions on PersonQA, an internal benchmark used by OpenAI to evaluate a model’s knowledge about people.

Microsoft plans to extend the availability of this model to macOS, as well as other devices, although specific details have not been disclosed. Moreover, the company is making both gpt-oss-20b and its older counterpart, gpt-oss-120b, available via its hosted Azure AI Foundry platform.

Both new models are also accessible on Amazon’s AWS.