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

In Brief

If you’ve ever recovered yourself wishing that ChatGPT was a small spot quicker connected the uptake, OpenAI seems to beryllium answering your prayers.

The AI laboratory has rolled retired a caller mode called Ultrafast, which it says is designed to earnestly accelerate the gait astatine which its latest and astir almighty model, GPT 5.6 Sol, accomplishes its work.

The institution says that Ultrafast tin enactment astatine 14x the velocity of modular processing, delivering up to 750 output tokens — specified tokens correspond the chiseled pieces of substance generated by an LLM erstwhile it interacts with a quality — per second.

“Until now, getting real-time velocity typically meant choosing a smaller oregon much specialized model,” the institution said in a blog post connected Thursday. “Ultrafast points to advancement successful a caller direction: more utile enactment per second.”

OpenAI’s competitors, similar Anthropic, person likewise launched accelerated versions of their models. Claude has accelerated mode, though it doesn’t present the benignant of velocity that OpenAI is offering here.

OpenAI suggests that this high-octane mentation of GPT 5.6 Sol tin beryllium deployed crossed a fig of antithetic firm workflows, astir notably incidental response, lawsuit work and support, fiscal marketplace analysis, and e-commerce, among different applicable areas.

Ultrafast, which is presently being released successful preview, is being powered by OpenAI’s concern with chipmaker Cerebras. Currently, that preview is lone being made disposable to a tiny radical of customers, though OpenAI says that it volition grow entree to the diagnostic arsenic “capacity grows.”

Subscribe for the industry’s biggest tech news