OpenAI took a strong step to attract corporate clients by introducing Ultrafast, a new mode designed for its most powerful model, GPT-5.6 Sol, to operate at a much higher speed. According to the company, this option allows for speeds of up to 14 times the standard processing speed and can generate up to 750 output tokens per second.
The proposal aims to address a common complaint among users of such systems: the delay in responses. Until now, achieving speeds close to real-time often meant sacrificing power and opting for smaller or specialized models. Ultrafast seeks to change that logic and demonstrate that it is possible to deliver more useful work per second without sacrificing capacity.
In the statement released this Thursday, the company emphasized this idea: “Until now, achieving real-time speed usually meant choosing a smaller or more specialized model. Ultrafast aims for progress in a new direction: more useful work per second.”
The mode is specifically designed for business workflows. Among the uses that OpenAI highlights are incident response, customer service and support, financial market analysis, and e-commerce, although the list of potential applications is broader.
