OpenAI is launching a preview of Ultrafast, a new service tier that runs the GPT-5.6 Sol model up to 14 times faster than Standard processing. Powered by Cerebras hardware, this mode generates up to 750 output tokens per second and is initially available via the OpenAI API.
The service targets time-sensitive workflows where real-time intelligence is critical, including:
- Incident response and reliability for analyzing logs during outages.
- Financial research and security for assessing changing market signals.
- Customer support and voice for resolving complex issues without interrupting conversations.
- Commerce applications like personalized recommendations and checkout assistance.
- Live research and experimentation to enable interactive working sessions.
OpenAI states that Ultrafast allows businesses to build more responsive products and make faster decisions by bringing frontier intelligence into demanding workflows without sacrificing speed.