On August 13, 2026, OpenAI published an early look at Ultrafast, a new service tier that runs GPT‑5.6 Sol up to 14× faster than Standard processing. The company says the mode can generate as many as 750 output tokens per second and is powered by hardware from Cerebras.
What Ultrafast offers
Ultrafast is presented as a capability intended for workflows where latency matters and where developers previously had to choose smaller or more specialized models to meet speed requirements. OpenAI describes the change as delivering more work per second without sacrificing the intelligence of its frontier model, GPT‑5.6 Sol.
Inside the company, teams have begun testing Ultrafast on tasks such as rapidly reading logs, analyzing traces, synthesizing conversations and accelerating research loops that historically ran as overnight batches. OpenAI states these uses shorten the delay between observing a signal, testing a hypothesis and selecting the next action, while leaving engineering judgment and deployment decisions to people.
Early customer use cases
OpenAI says an initial group of customers across coding, commerce, finance and support are trialing Ultrafast in production environments to identify where higher throughput brings the most value. Reported scenarios include incident response during outages, real-time financial research, voice-driven customer support, commerce interactions and interactive experimentation.
Industry contacts quoted by OpenAI described practical changes. John Crepezzi of Jane Street said the speed increase from Cerebras enables new ways of using the models and makes developers more productive. Courtland Lykins of Podium said the technology materially changed complex call experiences in their voice stack. Mitch Troyanovsky of Basis said Ultrafast lets teams create synchronous user experiences that were previously limited by intelligence, and Alex Wang at Rogo said it makes certain financial research feel like a real-time interaction.
Availability and next steps
OpenAI characterizes Ultrafast as the next phase of its partnership with Cerebras and confirms that GPT‑5.6 Sol on Ultrafast mode is available in a limited preview to a select group of customers. The company says it will expand access as capacity grows and invites businesses that require high-speed frontier intelligence to sign up for notifications when access widens.
Original source: OpenAI News