OpenAI has introduced a new mode called Ultrafast, which it states significantly accelerates its latest model, GPT 5.6 Sol. The company claims this new mode can operate at 14 times the speed of standard processing, delivering up to 750 output tokens per second.
According to OpenAI, this development allows for "more useful work per second," contrasting with previous methods where real-time speed often required using smaller or more specialised models. Competitors, such as Anthropic with its Claude fast mode, have also launched accelerated versions of their models, though OpenAI suggests its offering provides greater speed.
OpenAI proposes that this high-speed version of GPT 5.6 Sol could be used in various corporate applications, including incident response, customer service, financial market analysis, and e-commerce. The Ultrafast mode is currently in a preview phase, accessible to a limited number of customers, and is supported by OpenAI's partnership with chipmaker Cerebras. OpenAI plans to expand access to the feature as its capacity increases.