Ultrafast Mode Announced for GPT-5.6 Sol in Collaboration with OpenAI and Cerebras

OpenAI has announced an early preview of its new Ultrafast service tier, powered by Cerebras infrastructure and offering speeds up to 14 times higher compared to the standard processing procedure. Initially launched via the OpenAI API, this new system can generate up to 750 output tokens per second.

Features of the New Service Tier

Announced by OpenAI, the Ultrafast mode can boost the GPT-5.6 Sol model, one of the company's smartest models, up to 14 times the standard processing speed. Powered by Cerebras, this new tier draws attention with its capacity to generate 750 output tokens per second.

Where Speed Meets Intelligence

In the past, achieving real-time speed typically required selecting smaller or more specialized models. The Ultrafast mode signals a new era where speed can be increased without compromising artificial intelligence intelligence.

Use Cases and Scenarios

Designed for operations requiring high speed, this system is expected to provide significant advantages in time-sensitive business processes such as incident response, financial research, customer support, e-commerce, and live research.

Preview Period and Access Status

During the preview period of the system, OpenAI is working closely with an initial customer group to identify areas where speed makes the biggest difference. The GPT-5.6 Sol Ultrafast mode has been made available as a limited preview to expand as capacity increases.