OpenAI Introduces Ultrafast Mode for GPT-5.6 Sol.. Speeds Up AI by 14x

OpenAI has introduced an early preview of Ultrafast, a new API service tier designed to significantly reduce response times for GPT-5.6 Sol. The company says GPT-5.6 Sol on Ultrafast can generate up to 750 output tokens per second using Cerebras infrastructure and can run up to 14 times faster than standard processing. The initial preview is limited to select customers while OpenAI evaluates performance and gathers feedback.

OpenAI

The company says the faster processing could benefit real-time incident response, market-signal analysis, transaction assessment, customer support, live research and other interactive workflows. OpenAI is also testing the technology across coding, commerce, financial research and customer-support applications. The company plans to expand access as capacity increases and use insights from the preview to guide wider deployment.

OpenAI