How Google’s New Gemini Rates Work and How to Track Your Usage
Google has revamped its Gemini AI usage metering, shifting from counting requests to measuring the computing power required for prompts, impacting users across Free, Plus, Pro, and Ultra tiers.
Intelligence analysis by Gemini 2.5 Flash
Google's Gemini AI now meters usage based on the computational resources consumed by requests rather than a simple count of prompts, making it harder for users to predict when they'll hit limits. This change, designed to align with Google's operational costs, affects all subscription tiers and is influenced by prompt complexity, model choice, and context window size.
Imagine Google's AI is like a super-smart helper, but it uses special energy. Before, Google just counted how many times you asked the helper to do something. Now, it counts how much energy the helper uses for each task. So, asking for a simple weather report uses less energy than asking it to write a whole story or draw a picture. If you use too much energy, you have to wait a bit before the helper can work again, like waiting for a battery to recharge.
Analysis
The Shift to Resource-Based Metering
Google has fundamentally altered how it measures and meters usage for its Gemini AI applications, moving away from a simple count of requests to a more complex system based on the computing power required for each prompt. This means that instead of a fixed number of image generations or video creations per day, users now face limits determined by the complexity and resource intensity of their requests. From Google's perspective, this change offers a more accurate reflection of the actual operational costs incurred by its data centers, aligning usage charges with the computational load.
However, for end users, this new approach introduces a significant degree of vagueness. The article highlights that users can no longer rely on straightforward rules like "five image generations a day," as a complex request might consume more credits than a simpler one. This unpredictability can lead to frustration, with users unexpectedly hitting their limits and being told to wait before making further requests. Google's support documentation further complicates matters by stating that access is subject to change based on testing, experimentation, or availability, implying that daily limits may fluctuate without explicit notice.
Tiered Access and Model Capabilities
Google offers four main Gemini AI subscription tiers in the US: Free, AI Plus ($8/month), AI Pro ($20/month), and AI Ultra ($100 or $200/month). Each tier provides progressively higher usage limits, described as multipliers of the "standard" free tier limits: 2x for Plus, 4x for Pro, and 5x or 20x for Ultra, depending on the specific Ultra plan. All users have access to various Gemini AI models, including Flash-Lite, Flash, and Pro, with smarter models consuming more usage credits.
Beyond the base models, Google also introduces different "thinking" levels—Standard, Extended, and Deep Think—which influence response quality, speed, and, consequently, usage. Another critical differentiator across plans is the context window size, which dictates how much information can be included in a single conversation. Free users are limited to 32K tokens (approximately 24,000 words), AI Plus users get 128K tokens (about 96,000 words), and AI Pro and Ultra users benefit from a substantial one million tokens (around 750,000 words), enabling much longer and more complex interactions.
Monitoring Your AI Consumption
Despite the increased complexity of the new metering system, Google has provided tools for users to track their AI usage. Within the Gemini web app, users can navigate to the cog icon and then "Usage limits" to view their current consumption. Mobile app users can find this information via the menu button, then the cog, and finally "Usage limits." This screen displays two progress bars: one for current usage, which resets every five hours, and another for the weekly limit.
If a user exhausts their current usage, the Gemini app will indicate the time until the next reset, requiring them to wait before continuing. Hitting the weekly limit on a paid plan results in a temporary demotion to the most basic AI model until the next weekly reset. The article also notes that Google's support documents warn that limits may change without notice due to capacity constraints, with free users potentially being affected first. This transparency, while helpful for tracking, underscores the dynamic nature of AI resource allocation and the potential for unexpected service adjustments.
Key points
- Google's Gemini AI now meters usage based on computing power required for requests, not just the number of requests.
- Usage limits are influenced by the user's plan (Free, Plus, Pro, Ultra), prompt complexity, chosen AI model, and 'thinking' level.
- Context window size, indicating how much information can be processed in a conversation, varies significantly across plans.
- Users can track their current and weekly AI usage limits within the Gemini app, with resets occurring every five hours and weekly, respectively.
- Hitting limits on paid plans results in temporary demotion to the most basic AI model until the next reset.
Google's shift to resource-based metering could lead to more efficient allocation of its powerful AI infrastructure, ensuring that paying users receive consistent access to advanced features. For users, understanding the tiered system and context window sizes allows for better planning and budgeting of their AI usage based on specific needs.
The new, less transparent usage metering system makes it difficult for users, especially those on the free tier, to predict when they will hit their limits, leading to unexpected interruptions and frustration. Google's policy of potentially changing limits without notice further introduces uncertainty regarding service availability and consistency.
