OpenAI released gpt-oss-120b and gpt-oss-20b, its first open-weight models since GPT-2, available on Azure AI Foundry and Windows AI Foundry as of August 2025. The 120b model targets reasoning workloads at o4-mini performance level on single datacenter GPUs; the 20b targets agentic tasks on local hardware (Windows laptops with 16GB+ VRAM). Both will support API compatibility with OpenAI's responses API, enabling drop-in replacement into existing applications. This materially expands the set of frontier-capable open-weight models available to enterprises and developers, enabling fine-tuning, distillation, and on-device inference at the edge.
315 other tracked developments touching the same claims.