OpenAI unveils Ultrafast mode for GPT-5.6 Sol, up to 14x faster

OpenAI's new Ultrafast tier represents a strategic move to decouple speed from model size, addressing a common trade-off in AI deployment. By powering their most intelligent model at the time of the announcement with Cerebras' ultra-low-latency inference, OpenAI is targeting the most time-sensitive business processes, from incident response to live customer interactions. This builds on an existing partnership with Cerebras, signaling a deepening collaboration to push the limits of inference speed.

The company is testing the service with an initial group of customers across coding, commerce, and financial research, aiming to learn how order-of-magnitude speed changes product design and value creation. Inside OpenAI, teams are already using Ultrafast to compress overnight experiment loops into multiple daytime iterations and to accelerate incident response.

The stakes are clear: if Ultrafast delivers on its promise, it could redefine what constitutes a "real-time" AI application, shifting the focus from raw intelligence to throughput. However, the preview's limited availability means the outcome hinges on how the technology performs in real production environments and how effectively OpenAI scales access to meet demand. For businesses, the key watchpoint is whether the speed gains translate into measurable improvements in decision-making and customer experience as access broadens.