GPT-5.6 Sol Ultrafast Mode Preview: Up to 14x Faster Speed

OpenAI has introduced a new service tier called Ultrafast. This mode runs GPT-5.6 Sol up to 14x faster than the existing Standard processing and is initially launched via the OpenAI API.

Powered by Cerebras technology, Ultrafast can generate up to 750 output tokens per second. This provides revolutionary speed for business workflows where real-time performance is essential, without sacrificing the model's intelligence.

Key use cases include:

• Incident Response and Reliability: Instantly analyzing logs and code changes during system failures to help find solutions.
• Financial Research and Security: Analyzing real-time market signals and identifying suspicious transactions.
• Customer Support and Voice: Resolving complex customer issues in real time without interrupting conversation flow.
• Commerce: Answering real-time product questions and checking inventory during shopping.
• Real-time Research and Experimentation: Accelerating experimentation by turning research processes into interactive sessions.

Early customers such as Jane Street, Podium, Basis, and Rogo are currently testing the service, praising the breakthrough speed improvements while maintaining the model's intelligence.