GPT-6 Astra Ultrafast: up to 8× faster on Blackwell

NVIDIA’s blog says GPT-6 Astra Ultrafast is available now in the OpenAI API and for eligible ChatGPT Work and Codex users, running on NVIDIA Blackwell GPUs. OpenAI reports up to 8× faster token generation than Astra Standard mode, aimed at shorter edit–test–debug loops, less wait between tool calls, and snappier interactive apps. Quotes from OpenAI inference and compute leads stress using their models to optimise kernels on Blackwell/Rubin and keep improving inference after deploy. No separate discuss_url in the pick list; treat as lab post. Sorted take: Latency-sensitive coding agents should A/B Standard vs Ultrafast on cost per successful task—not raw tokens—before switching defaults.