OpenAI preview Ultrafast service GPT-5.6 Sol speed up to 14 times

OpenAI is previewing a new way to run its most powerful GPT-5.6 model at higher speeds. The company says the new Ultrafast service tier allows the GPT-5.6 Sol to run up to 14 times faster than standard processing. How it works Ultrafast mode is first available through the OpenAI API and is powered by Cerebras. It can generate up to 750 output tokens per second, with the potential to bring cutting-edge performance to workflows where latency is as important as model intelligence. Model Background OpenAI Launched in June as the GPT-5.6 Sol, the series also includes the balanced Terra and speed Luna. The entire model family was made generally available in July, including through ChatGPT, Codex and API. Applications and Limitations OpenAI considers Ultrafast to support real-time or near-production tasks, including voice, customer support, commerce, developer agency, financial research, and security response. The company says its own developers have used it to analyze logs and traces during incidents, compressing research cycles that once required overnight runs into multiple iterations during the workday. The service is currently limited to specific customers while OpenAI evaluates the changes brought by the new rate to the actual product and expands capacity. Enterprises can join the waitlist by sharing workload, latency requirements, expected usage and other information.