Cerebras Powers Ultrafast Mode in OpenAI API for GPT-5.6 Sol

Cerebras said it powered Ultrafast mode, a new service tier in the OpenAI API for GPT-5.6 Sol. The company tied the change to higher output speeds and lower latency for users starting with a limited preview.

Ultrafast runs GPT-5.6 Sol at up to 750 output tokens per second, with performance described as up to 14× faster than Standard processing. Cerebras said Ultrafast provides the same intelligence as GPT-5.6 Sol Standard. The service started as a limited preview for OpenAI customers.

Cerebras said Ultrafast’s speed comes from its Wafer-Scale Engine architecture, which keeps model weights on-chip at 44 GB of SRAM per wafer-sized chip. The company said that approach avoids shuttling weights between on-chip memory and off-chip storage, which it said eliminates a memory-bandwidth bottleneck found in GPU-based inference.

Cerebras said it powers the Ultrafast tier and that OpenAI is combining GPT-5.6 Sol with Cerebras’ inference technology. OpenAI said it started with a small group of customers to learn where speed creates value, and that it would use those learnings to inform expansion over time.

“GPT-5.6 Sol on Ultrafast is proof that speed and intelligence are no longer mutually exclusive,” said Andrew Feldman, CEO and co-founder, Cerebras. “Together with OpenAI, we're putting frontier intelligence in the hands of users at unprecedented speed and changing what's possible with AI.” “By combining GPT-5.6 Sol with Cerebras’ inference technology, we’re exploring what becomes possible when customers can get the intelligence of our most capable models with significantly lower latency. We’re starting with a small group of customers to learn where that speed creates meaningful value, and we’ll use those learnings to inform how we expand the service over time,” said Sachin Katti, VP Compute Strategy & GPT-Infra at OpenAI.

Provided by Globe Newswire on behalf of Cerebras. Click to read original content. The original article was written by Decision Insights Editorial.