A model rollout is the phased release of a new AI model to users. Companies often stage these releases across different tiers—such as free users, paying subscribers, and API users—to manage computational resources and gather feedback.

The recent launch of OpenAI's GPT-6 Astra, which locked out many paying users, demonstrates how capacity constraints and unexpected demand can cause significant access problems during a rollout. Even with careful planning, the sheer scale of modern AI models and the infrastructure needed to run them can lead to outages if demand outstrips the available computing power.