OpenAI introduces ‘Ultrafast,’ a new mode that makes GPT-5.6 Sol work at 14x the speed

OpenAI has launched a new mode called Ultrafast that accelerates its GPT-5.6 Sol model by fourteen times. This update allows the system to generate up to 750 output tokens per second, moving beyond the usual trade-off between speed and intelligence. Previously, achieving real-time responses required switching to smaller or more specialised models. The company states that Ultrafast enables more useful work per second without sacrificing the capabilities of the latest large language model. Competitors such as Anthropic have introduced similar fast modes, but none currently match the velocity offered here. OpenAI indicates that this high-performance version is suitable for corporate workflows including incident response, customer service, financial market analysis, and e-commerce. The feature relies on a partnership with chipmaker Cerebras and is currently in preview. Access is limited to a small group of customers while capacity expands. Ultrafast represents a significant shift in how organisations deploy generative AI for time-sensitive tasks.