Microsoft brings OpenAI's new open model to Windows for local AI

Microsoft has announced the immediate availability of OpenAI’s new lightweight, open-source GPT model, gpt-oss-20b, for Windows users. This marks the first time an OpenAI model can be run natively on the Windows operating system, with support for macOS also slated for release soon through Microsoft’s local AI development initiatives.

The gpt-oss-20b model, which OpenAI recently released as a free and open-source offering, is designed for local execution directly on personal computers. To run the model, Windows users will require a PC or laptop equipped with at least 16GB of VRAM, typically found in high-end GPUs from manufacturers like Nvidia or AMD Radeon.

Microsoft has integrated gpt-oss-20b into its Windows AI Foundry platform, pre-optimizing it for efficient local inference. The company highlights the model’s suitability for specialized tasks, stating it is “perfect for building autonomous assistants or embedding AI into real-world workflows, even in bandwidth-constrained environments.” Its optimization for code execution and tool use makes it a versatile asset for developers and advanced users.

This move by Microsoft suggests a broader strategy to enhance local AI capabilities within Windows. The company has hinted at future support for a wider range of devices, potentially including more optimized versions for its Copilot Plus PCs, as it continues to integrate various local AI models into the operating system. Running AI models locally offers significant advantages, including enhanced privacy by keeping data on-device, reduced latency due to no cloud dependency, and improved functionality in areas with limited internet access.

Microsoft’s swift adoption of OpenAI’s latest open model for Windows mirrors Amazon’s equally rapid integration of the new open-weight GPT-OSS models into its cloud services. This parallel development adds another layer of complexity to the already intricate partnership between OpenAI and Microsoft. While Microsoft has invested heavily in OpenAI and tightly integrated its models into its cloud and product ecosystems, OpenAI’s decision to release open-weight models, and their quick adoption by competitors, signals a shift towards broader accessibility of advanced AI models across the industry.