AI
Google's Gemini AI Now Charges by Computing Power, Not Request Count
Google has overhauled how it tracks AI usage across Gemini's free and paid plans, switching from simple request limits to a more complex resource-based system.
WIRED — AI
Google has fundamentally changed the way it meters Gemini AI usage, shifting from counting individual requests to measuring the actual computational resources your prompts consume. This overhaul affects all subscription tiers—Free, Plus, Pro, and Ultra—and means users now face less predictable spending patterns.
The New Resource-Based Model#
Under the old system, limitations were straightforward: you might generate three videos daily. Today, two video generations could exhaust your quota if they're particularly demanding. Google's logic is sound from an infrastructure standpoint—charging by actual server load rather than request volume reflects true operational costs. For users, though, this creates uncertainty. You can no longer follow a simple rule like "5 image generations per day" because complexity is now the determining factor.
Complicating matters further, Google reserves the right to adjust access "based on testing, experimentation or availability." Translation: your available capacity may fluctuate without warning.
What Drives Your Usage Limits#
Three elements combine to determine how much AI capacity you can consume:
- Your subscription plan (free, Plus, Pro, Ultra)
- Prompt complexity and length (a weather request versus custom code generation)
- Which Gemini model you select (Flash-Lite, Flash, or Pro, each with Standard, Extended, and Deep Think thinking levels)
Each model consumes different amounts of your quota, and thinking levels add further granularity to resource consumption.
Tier-by-Tier Breakdown#
Google offers four paid Gemini tiers in the US alongside a free option. AI Plus costs $8/month, Pro runs $20/month, and Ultra comes at either $100 or $200 monthly. Each paid tier unlocks multiples of the baseline free limits: Plus grants 2x, Pro gives 4x, and Ultra provides either 5x or 20x depending on which Ultra option you choose.
Free users get a 32K-token context window (roughly 24,000 words per conversation), Plus users reach 128K tokens (96,000 words), and Pro/Ultra users unlock up to a million tokens (750,000 words). Larger context windows let you feed more background information into a single thread.
