OpenAI introduces ‘Ultrafast,’ a new mode that makes GPT-5.6 Sol work at 14x the speed

OpenAI has unveiled a new mode called Ultrafast, which promises to significantly speed up the performance of its powerful model, GPT-5.6 Sol. The AI company claims that Ultrafast can work at an astonishing 14 times the speed of standard processing, delivering up to 750 output tokens per second. Output tokens are individual units of text generated by a large language model when it interacts with a user.

Traditionally, real-time speed typically required selecting a smaller or more specialized model. However, Ultrafast marks a new direction for OpenAI, focusing on generating more useful work per second. Competitors such as Anthropic have also introduced accelerated versions of their models, with Anthropic's Claude offering a fast mode, though it doesn't match the speed offered by OpenAI.

OpenAI envisions deploying Ultrafast across various corporate workflows, particularly in incident response, customer service and support, financial market analysis, and e-commerce. The new mode is currently in preview and is powered by OpenAI's partnership with chipmaker Cerebras. At present, this preview is only accessible to a select group of customers, but OpenAI plans to expand availability as capacity grows.