OpenAI’s GPT-6 Astra Ultrafast is available now in the OpenAI API and to eligible ChatGPT Work and Codex users, running on NVIDIA Blackwell GPUs. NVIDIA describes Ultrafast as leveraging inference optimizations that tap into Blackwell’s capabilities to deliver substantially faster token generation than Astra Standard mode.
AVAILABILITY AND PERFORMANCE GAINS
Ultrafast offers up to 8x faster token generation than Astra Standard, according to NVIDIA’s briefing. This acceleration is designed to shorten coding agents’ edit-test-debug cycles, reduce the time spent generating responses between tool calls, and make interactive applications feel more responsive. TechStaged has also covered OpenAI Adds More Admin Control for ChatGPT Work and Codex.
- Available now via the OpenAI API and to eligible ChatGPT Work and Codex users.
- Optimization targets latency, throughput, and cost across typical developer workflows.
WHY THE SPEED MATTERS FOR DEVELOPERS
Faster generation matters most when the result is repeated across a workflow—agents write code, use tools, check results, and decide what to do next. Ultrafast is intended to bring Astra’s capabilities into these time-sensitive loops.
- Speeds up interactive tasks
- Reduces wait times between successive actions in a workflow
VOICES FROM OPENAI AND NVIDIA
NVIDIA says its tooling and documentation helped OpenAI make models performant on Blackwell and Rubin GPUs. Astra Ultrafast translates that knowledge into high-performance kernels that improve latency, throughput, and cost across workloads.
- “NVIDIA’s deep investment in tooling and documentation has enabled us to make our models exceptionally good at programming Blackwell and Rubin GPUs.” — Philippe Tillet, OpenAI
WHAT HAPPENS NEXT FOR ASTRA ULTRAFAST
OpenAI notes ongoing work to refine inference software using its own models on NVIDIA GPUs, aided by NVIDIA’s programmable platform. This compatibility lets developers reuse infrastructure across training, inference, and reinforcement learning as models evolve, improving utilization and adaptability.
- Internal models are used to optimize inference on NVIDIA GPUs.
- Programmability supports evolving workloads and resource reuse.
RELATED COVERAGE
- OpenAI Adds More Admin Control for ChatGPT Work and Codex
- Stampli accelerates product launch using Codex and ChatGPT Work to go from weeks to days
- NVIDIA scales expertise by deploying ChatGPT Work to automate tasks and accelerate workflows
- Polimill powers Japan’s next-generation public AI infrastructure with GPT models and Codex
- AI & Automation articles
SOURCES
- NVIDIA Blog: How NVIDIA GPUs Help Accelerate OpenAI’s GPT-6 Astra Ultrafast Published · Primary source
