How NVIDIA GPUs Help Accelerate OpenAI’s GPT-6 Astra Ultrafast
nvidia ·
OpenAI has released GPT-6 Astra Ultrafast, a new model mode available via the OpenAI API and for eligible ChatGPT Work and Codex users. This version runs on NVIDIA Blackwell GPUs and delivers up to eight times faster token generation compared to the standard Astra mode. The acceleration is achieved through continuous inference optimizations that leverage the capabilities of the NVIDIA Blackwell architecture. These speed improvements are designed to shorten coding agents' edit-test-debug cycles and reduce latency between tool calls in interactive applications. Philippe Tillet, an inference lead at OpenAI, highlighted how NVIDIA's tooling enabled their models to create high-performance kernels for NVIDIA hardware. Uday Ruddarraju, OpenAI's chief technology officer of compute, noted that the company used its internal models to optimize inference software on NVIDIA's programmable platform. This collaboration allows developers to repurpose infrastructure across training, inference, and reinforcement learning as demand shifts. Developers can currently access GPT-6 Astra Ultrafast through the API, with further details available in the official guide.
nvidia