Google Gemini Now Metered by Compute, Not Requests
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Google revamped Gemini usage limits so they're now measured by the computing power a prompt requires rather than the number of requests, making it harder for users to predict when they'll hit a cap.
- Gemini offers four US subscription tiers — Free, AI Plus ($8/month), AI Pro ($20/month), and AI Ultra ($100 or $200/month) — with paid plans delivering 2x to 20x the free tier's unspecified "standard" limits.
- All Gemini users can access the Flash-Lite, Flash, and Pro models along with Standard, Extended, and Deep Think reasoning levels, with smarter models and deeper reasoning consuming more of the user's quota.
- Gemini's context window scales sharply by tier: 32K tokens (~24,000 words) for free users, 128K (~96,000 words) for AI Plus, and 1 million tokens (~750,000 words) for AI Pro and Ultra subscribers.
- Gemini usage resets on a 5-hour rolling window and a weekly window, viewable in the app's settings under Usage limits; paid users who exhaust their quota are demoted to the most basic model until the next reset.
- Google's support docs state that limits may change without notice due to capacity constraints, with free users the first to feel any throttling when resources tighten.
Why it matters: Google's switch to compute-based metering means heavy users of advanced Gemini features (video generation, coding, Deep Think reasoning) will burn through quotas faster, even if they're sending fewer prompts. Paid subscribers at the $20/month AI Pro tier are paying for 1M-token context windows but get demoted to the base model if they exceed limits, effectively turning the subscription into a soft cap rather than unlimited access.
