Gemini usage limits go live

Google’s new compute-based usage limits for Gemini are now live, and users began publicly complaining this week after long-running chats and agent tasks were cut off. 9to5Google reported on May 21 that the changes, announced around Google I/O 2026, had triggered frustration among existing users citing interrupted sessions and tighter caps. Google’s own support documentation says the new system began on May 17, 2026, replacing the older, simpler prompt-style limits with a broader compute-based model. The company says the limits depend on the complexity of a prompt, the model or feature used, and the length of a chat. ### When did the new Gemini caps actually start? Google’s Gemini Apps Help page says “Starting on May 17, 2026 there will be changes to usage limits for Gemini Apps.” The page describes the new framework as compute-based, meaning access is not measured only by a raw number of prompts but by how resource-intensive a session is. May 21 was the date the change became widely visible to users following Google I/O coverage and public complaints. 9to5Google said limits had “not really been much of a thought for Gemini users to date,” but that changed after the new system went live and users started posting about interrupted use. ### What does “compute-based” mean in practice? Google’s support page says Gemini usage now depends on “the complexity of your prompt, the models and features you use, and the length of your chat.” The company also says a user’s limit “refreshes every 5 hours until you reach your weekly limit.” That means a short exchange and a long, tool-heavy session do not count the same way. Google does not provide a simple universal prompt number on the support page; instead, it says usage can vary by feature and by plan, including free access, Google AI Pro and Google AI Ultra. ### Which Gemini features appear to be affected? 9to5Google reported that users were reacting to limits affecting Gemini’s newer tools and longer-running workflows, including continuous-agent style use. The publication cited complaints from users who said sessions were being stopped or constrained after the caps appeared. Google’s broader Gemini updates page this week highlighted features such as Gemini 3 Pro, Gemini Omni, Daily Brief and agent-style capabilities, while Google One support pages say AI Ultra subscribers get “the highest usage limits” and prioritized traffic for Antigravity, Google’s Gemini-powered development environment. Those support documents indicate the new cap structure applies across the expanding Gemini app ecosystem, though access levels differ by subscription tier. ### Did Google change anything after users pushed back? 9to5Google reported on May 21 that Google had already raised limits in one area, saying usage limits for Antigravity had been increased twice after the caps took effect. The report said user frustration had been strongest around the new restrictions, particularly for people testing more intensive workflows. Google’s public help pages still describe the main system as a rolling refresh every five hours with an overall weekly ceiling. The company has not, in the support documents reviewed, published a single public chart spelling out exact compute allotments for every Gemini feature. ### Why are users reacting now instead of at I/O? May 19 and May 21 were the key dates in public coverage. Google’s product rollout around I/O introduced new Gemini models and agent features, but the user response intensified once people hit actual limits in live use rather than reading about them in announcements. 9to5Google’s May 21 report captured that shift from launch messaging to user experience. Google’s Help Center remains the main public source for the mechanics of the limits, and the page says the changes started May 17, with refreshed usage windows every five hours and a weekly cap still in place.