OpenAI announced in August 2026 a new processing mode called Ultrafast, designed to dramatically increase the speed of its latest model, GPT-5.6 Sol. The company says Ultrafast can operate at 14 times the speed of standard processing, delivering up to 750 output tokens per second.
The feature is currently available in preview to a limited group of customers, with OpenAI stating it will expand access as “capacity grows.” Ultrafast is powered by OpenAI’s existing partnership with chipmaker Cerebras.
OpenAI is positioning Ultrafast primarily for enterprise use cases, citing incident response, customer service and support, financial market analysis, and e-commerce as areas where the accelerated model could be deployed.
“Until now, getting real-time speed typically meant choosing a smaller or more specialized model,” OpenAI said in a blog post. “Ultrafast points to progress in a new direction: more useful work per second.”
The move appears aimed at attracting enterprise customers who require high-throughput AI processing. OpenAI’s competitor Anthropic has also released a fast mode for its Claude model, though OpenAI claims Ultrafast delivers greater speeds than what Claude’s fast mode offers.
For context, output tokens represent the distinct pieces of text an AI language model generates when responding to a user. Higher token-per-second rates could mean faster response times in applications where speed is critical, such as real-time customer interactions or time-sensitive financial analysis.
Source: TechCrunch