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

OpenAI has started previewing Ultrafast, a new mode that it says lets its latest and most powerful model, GPT-5.6 Sol, work at roughly 14x the speed of standard processing.

According to TechCrunch's report, which cites OpenAI's Thursday blog post, Ultrafast can deliver up to 750 output tokens per second. Until now, getting real-time speed typically meant choosing a smaller or more specialized model, the company said, and Ultrafast points to progress in a new direction: more useful work per second.

The mode is powered by OpenAI's partnership with chipmaker Cerebras, and the preview is currently available only to a small group of customers. OpenAI says it will expand access to the feature as capacity grows.

OpenAI suggests the high-speed variant can be deployed across corporate workflows including incident response, customer service and support, financial market analysis, and e-commerce.

The move is aimed at courting enterprise users, who often need low-latency responses for real-time operations.

Competitors have pursued similar paths, and Anthropic's Claude has a fast mode, but TechCrunch notes it does not deliver the kind of speed OpenAI is offering here.

The release is significant because real-time speed has typically meant trading down to smaller or specialized models; Ultrafast claims frontier-level capability at 14x throughput, pointing to a future where latency no longer forces a capability trade-off.

What to watch next: when Ultrafast moves from preview to general availability, how it is priced, and whether Cerebras-powered speed becomes a standard feature of frontier models.