UMVA has learned that Google is overhauling the way it measures weekly Gemini usage, shifting from a simple request count to a dynamic, compute‑driven model.
This new approach weighs the intricacy of each prompt, the heavy‑hitting features you tap—such as image and video generation, deep research, or the Pro and Deep Think models—and even the length of your conversation.
While the exact formulas remain cloaked in mystery, the company confirms that paying subscribers will enjoy far loftier ceilings than free users.
Subscribers to the $8‑a‑month AI Plus tier now receive twice the standard allowance, the $20‑a‑month AI Pro tier quadruples it, and the premium $250‑a‑month AI Ultra plan rockets the limit to twenty times the baseline.
These compute‑based quotas reset every five hours, ticking down until the weekly cap is reached.
Until now, Gemini users were bound by a flat daily request limit—AI Pro members could fire off up to 100 prompts a day, regardless of how complex those prompts became.
The timing is striking: less than a month after GitHub scrapped its “premium request units” in favor of token‑based AI Credits, Google is following suit to tame the voracious appetite of ever‑more autonomous AI features.
Powerful agentic tools can spawn sub‑agents that devour tens of thousands of tokens across multiple turns, threatening to outpace traditional flat‑rate plans.
Even rivals are feeling the pressure; Anthropic recently doubled its Claude Code limits after securing a massive compute boost from a new partnership, acknowledging that its existing plans weren’t built for the latest agent‑driven capabilities.
