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

OpenAI is addressing the demand for high-performance AI with the launch of Ultrafast, a new operational mode designed to significantly boost the performance of its flagship model, GPT-5.6 Sol. By prioritizing raw speed, the company aims to make its most capable technology more viable for demanding enterprise environments.

Unprecedented Throughput

The new mode is engineered to operate at 14 times the speed of standard processing. According to OpenAI, this allows the model to generate up to 750 output tokens per second, a massive leap in efficiency for large language model interactions.

“Until now, getting real-time speed typically meant choosing a smaller or more specialized model. Ultrafast points to progress in a new direction: more useful work per second.”

Strategic Enterprise Applications

While competitors like Anthropic have introduced their own accelerated modes, OpenAI’s latest offering positions itself as a high-octane solution for complex corporate workflows. The company highlights several key areas where this speed will prove transformative:

• Incident Response: Rapid analysis of security or operational threats.
• Customer Support: Instantaneous, high-volume query resolution.
• Financial Analysis: Real-time processing of market data.
• E-commerce: Dynamic, high-speed interaction for shopping platforms.

Partnership and Availability

The performance gains behind Ultrafast are made possible through a strategic collaboration with chipmaker Cerebras.

Currently, the feature is available in a limited preview phase for a select group of customers. OpenAI has indicated that it intends to broaden access to the tool as its infrastructure capacity scales. This move marks a significant shift in how the lab balances the power of its largest models with the practical, real-time needs of the business world.