OpenAI Launches 'Ultrafast' Mode to Slash GPT-5.6 Sol Latency
AI-generated image · Bay Street Wire
The new mode, powered by a Cerebras partnership, targets high-volume corporate workflows with a 14x speed increase.
As first reported by TechCrunch, OpenAI has introduced "Ultrafast," a new operational mode designed to significantly accelerate the processing speed of its most powerful model, GPT-5.6 Sol. The Ultrafast mode can operate at 14x the speed of standard processing, capable of delivering up to 750 output tokens per second.
Historically, achieving real-time speed required enterprises to utilize smaller or more specialized models. OpenAI stated in a Thursday blog post that Ultrafast represents a shift toward achieving "more useful work per second." The company identified several high-volume corporate workflows where this speed is critical, including e-commerce, financial market analysis, customer service and support, and incident response.
TechCrunch reports that the technology is powered by a partnership with chipmaker Cerebras. While competitors such as Anthropic offer a "fast mode" for Claude, OpenAI claims its new offering provides superior speed.
Currently, Ultrafast is in a preview phase and is available to a limited group of customers. OpenAI noted that access to the feature will grow as capacity increases.