Microsoft brings OpenAI’s smallest open model to Windows users | TechCrunch

Microsoft is making OpenAI’s new free and open GPT model, gpt-oss-20b, available to Windows 11 users via Windows AI Foundry, the tech giant’s platform that lets users tap AI features, APIs, and popular open-source models on their computers.

The company said in a blog post that gpt‑oss-20b is “tool-savvy and lightweight,” adding: “Optimized for agentic tasks like code execution and tool use, it runs efficiently on a range of Windows hardware, with support for more devices coming soon. It’s perfect for building autonomous assistants or embedding AI into real-world workflows, even in bandwidth-constrained environments.”

Launched on Tuesday, OpenAI’s gpt-oss-20b can run on consumer PCs and laptops with at least 16GB of VRAM, which a modern GPU from Nvidia or Radeon would have. OpenAI said the model was trained using high-compute reinforcement learning, which helps it excel at powering AI agents and call tools, such as web search or Python code execution, as part of its chain-of-thought process.

However, the model is text-only, meaning it will not be able to process or generate images and audio like the company’s other models.

Also, the model has a habit of hallucinating a lot: gpt-oss-20b hallucinated in response to 53% of questions on PersonQA, OpenAI’s in-house benchmark for measuring the accuracy of a model’s knowledge about people.

Microsoft says it will bring the model to macOS soon, as well as other devices, though the company did not name any more. Microsoft is also making both gpt-oss-20b and older model gpt-oss-120b available via its hosted Azure AI Foundry platform.

The two new models are also available on Amazon’s AWS.