OpenAI's GPT-6 Astra Ultrafast Runs on NVIDIA Blackwell, Up to 8x Faster

What Ultrafast offers

NVIDIA says Ultrafast is accelerated by inference optimizations in OpenAI's models that take advantage of the NVIDIA Blackwell architecture. The result, according to the company, is up to 8x faster token generation compared with Astra Standard.

The mode is aimed at developers. NVIDIA argues that faster generation can shorten the edit-test-debug cycles of coding agents, cut the time spent generating responses between tool calls and make interactive applications feel more responsive.

The company stresses that speed matters most when it is repeated across a workflow. An agent writes code, uses a tool, checks the result and decides what to do next. Ultrafast, NVIDIA says, brings Astra's capabilities into these time-sensitive loops.

Models that optimize their own serving

Philippe Tillet, inference lead at OpenAI, said that "NVIDIA's deep investment in tooling and documentation has enabled us to make our models exceptionally good at programming Blackwell and Rubin GPUs." He added that Astra can turn that knowledge into high-performance kernels that make NVIDIA hardware compelling across latency, throughput and cost.

Uday Ruddarraju, chief technology officer of compute at OpenAI, said the company used its internal models to optimize inference on NVIDIA GPUs, and that NVIDIA's programmability helped deliver the acceleration behind Astra Ultrafast.

According to NVIDIA, this work continues after deployment. OpenAI is using its own models to refine the inference software running on NVIDIA GPUs, testing and implementing improvements. That ongoing effort can make model responses faster and deployed infrastructure more productive over time.

Flexibility across workloads and access for developers

NVIDIA also points to the programmability of its platform. It says developers and researchers can reuse infrastructure across training, inference and reinforcement learning as models evolve. That flexibility, the company says, helps teams repurpose compute as demand changes, improving utilization and avoiding overprovisioning for each workload.

Developers can use GPT-6 Astra Ultrafast through the API today. NVIDIA directs readers to the Ultrafast guide for access, pricing and implementation details. The announcement itself does not list prices or other specifics.

For teams building coding agents and interactive applications, the pitch is shorter waits at each step of a multi-step task. The 8x figure is a maximum stated by NVIDIA, so real-world gains will depend on the workload.