Cerebras and OpenAI preview Ultrafast API tier for GPT-5.6 Sol, claiming up to 750 output tokens per second

Cerebras and OpenAI introduced an early look at Ultrafast Mode, a new OpenAI API service tier initially available to select customers. The companies say GPT-5.6 Sol on Ultrafast can reach 750 output tokens per second without quality loss and completed a Humanity’s Last Exam run far faster than Claude Fable 5 in Cerebras’ testing.

The launch signals growing competition around inference speed as a premium differentiator, not just model quality. If the performance claims hold up in real-world workloads, faster frontier models could unlock more practical AI agents, coding tools, and time-sensitive enterprise workflows.