OpenAI announced Ultrafast on August 13 as a limited preview for running GPT-5.6 Sol more quickly through its API. A small group of customers receives initial access, with expansion dependent on additional capacity.
OpenAI reports up to 750 output tokens per second, or 14 times its standard processing speed. That measures output generation, not the time needed to finish an entire workflow.
Ultrafast runs through OpenAI's partnership with Cerebras. The announcement describes uses such as customer support and incident response, where response delay matters. It does not introduce an Ultrafast setting for every ChatGPT user.