Microsoft Adds New OpenAI Models to Azure and Windows

Microsoft is expanding its AI tools by bringing OpenAI's open-weight language models, gpt-oss, to Azure AI Foundry and Windows AI Foundry. Two of these models debuted this week: gpt‑oss-120b and gpt‑oss-20b. The larger 120b model is built for high-performance reasoning tasks, while the smaller 20b model works on PCs with GPUs that have at least 16GB of memory.

Developers have full access to the model weights, which means they can fine-tune the models to suit different needs. That could mean trimming for offline use or modifying them for industry-specific assistants. Open-weight access lets teams review models, retrain parts of them, or export them for use on Microsoft’s AKS or local computers.

Azure AI Foundry offers support via tools for model evaluation, fine-tuning, and deployment. It includes a growing catalog of more than 11,000 models. Meanwhile, Foundry Local offers on-device support, which can help those who ...