OpenAI has launched a new ultra-fast mode for its AI model as a preview. This mode, which produces 750 tokens per second, aims to reduce the time required for GPT-5.6 Sol to complete tasks. The company behind ChatGPT claims that the ultra-fast mode runs GPT-5.6 Sol much faster than standard processing. OpenAI has confirmed that it is working with a select group of customers to evaluate the performance of this new mode.

GPT-5.6 Sol gets an ultra-fast mode

According to "gadgets360", OpenAI announced an early preview of the ultra-fast mode in a blog post. The company stated that this new API service mode runs GPT-5.6 Sol up to 14 times faster than standard processing.

GPT-5.6 Sol operates in the ultra-fast mode using Cerebras, and it is claimed to produce up to 750 tokens per second. The company suggests that this will enable businesses to build more responsive products, make quicker decisions, and integrate powerful AI directly into their demanding workflows.

"GPT-5.6 Sol is available in ultra-fast mode today in a limited preview for a select group of customers," said OpenAI. The company has confirmed its plans to expand access as capacity increases.

OpenAI highlighted potential use cases for the ultra-fast speed, such as instant incident response, market indicator analysis, transaction evaluation, customer support, and direct research and experimentation.

The company confirmed that it is working with a “small group of customers” to “understand where speed makes a big difference, and how these insights can improve our products over time.”

Company tests GPT-5.6 Sol in high-speed mode

The company is testing GPT-5.6 Sol in high-speed mode at companies working in programming, trading, financial research, customer support, and other interactive applications.

The company says that the initial experiments help it understand where significant speed changes add value to workflows, adding “we will use these insights to guide our deployment as capacity grows.”

OpenAI developers are testing GPT-5.6 Sol internally in high-speed mode, and the team has confirmed its use of this mode in responding to incidents, with claims that this new mode reduces latency between signal detection, hypothesis testing, and choosing the next action.

The internal test also includes adopting the high-speed mode for research involving searching knowledge sources, querying data, and gathering, organizing, and summarizing information across connected tools.