Microsoft brings OpenAI’s new free GPT model into Windows 11 via AI

Microsoft has begun rolling out OpenAI’s newly released open-source language model, gpt-oss-20b, to Windows 11 users through its Windows AI Foundry platform. The move is part of Microsoft’s broader effort to bring powerful AI capabilities directly to users’ devices, reducing dependence on cloud-based infrastructure.

Read More: Satya Nadella says Microsoft layoffs ‘weighing heavily’ on him amid AI restructuring

The company stated that it aims to empower developers and users to access AI features, APIs, and leading open-source models locally on their PCs. Gpt-oss-20b, described as both “tool-savvy” and lightweight, is engineered for agentic tasks such as executing code and using digital tools autonomously. The model is optimised to perform efficiently across a variety of Windows hardware configurations, making it well-suited for building autonomous assistants or integrating AI into local workflows—especially in bandwidth-constrained environments.

Read More: Microsoft’s new AI agent independently detects and blocks malware

To run the model locally, however, systems must be equipped with GPUs offering at least 16GB of VRAM, typically found in recent Nvidia or AMD Radeon cards.

“The release of gpt‑oss and its integration into Azure and Windows is part of a bigger story. We envision a future where AI is ubiquitous—and we are committed to being an open platform to bring these innovative technologies to our customers, across all our data centres and devices,” Microsoft said in a blog post announcing the update.

Unlike OpenAI’s premium offerings such as GPT-4, gpt-oss-20b is strictly a text-only model. It does not process or generate images, audio, or video. Instead, it focuses on text generation and tool-calling functionality, having been trained using high-compute reinforcement learning to support tasks like web searches or executing Python scripts during reasoning sequences.

However, the model comes with notable limitations. On OpenAI’s internal PersonQA benchmark, gpt-oss-20b provided incorrect answers to questions about people 53% of the time, raising questions about its dependability in knowledge-intensive applications.

Looking ahead, Microsoft confirmed plans to expand support to macOS and additional hardware platforms. Both gpt-oss-20b and the larger gpt-oss-120b model will be made available via Azure AI Foundry and Amazon Web Services (AWS), extending their reach to more developers, businesses, and enterprise applications globally.