OpenAI has announced a new mode called Ultrafast for its latest flagship model, GPT-5.6 Sol, claiming a 14x speed increase over standard processing.

The mode can deliver up to 750 output tokens per second, enabling near-real-time generation.

The speed boost comes from a partnership with chipmaker Cerebras, though OpenAI did not disclose technical details about how the hardware accelerates inference.

Ultrafast is initially available in preview to a small set of customers, with broader access planned as capacity grows.

The company highlights use cases in incident response, customer service, financial market analysis, and e-commerce — domains where low latency matters.

The announcement positions Ultrafast as an alternative to using smaller, specialized models for speed, and notes that competitors like Anthropic have similar fast modes but at lower speeds.