跳至主要內容
Quick Access
AI Technology

Google Changes Gemini Limits: Switching from Daily Caps to Compute Usage

5 分鐘2 views

Here is the clean text version with all HTML tags removed:

Attention heavy AI users! Google has quietly introduced a major shift in how Google Gemini calculates its usage limits. The familiar "daily message count" is history. Starting May 17, 2026, a new policy based on "Compute Usage" has begun rolling out globally.

Core Change: From Message Count to Resource Allocation

Under the old system, quota tracking was simple. If you had a limit like "100 messages per day," every prompt counted as one request, whether it was a complex task or a simple hello. The new system, however, focuses on "how much AI computing resource each request consumes."

This means your usage will now vary dynamically based on complexity:

  • Lower Consumption: Short questions and basic text interactions.

  • Higher Consumption: Long conversations and analyzing massive text or code contexts.

  • Fastest Consumption: Advanced capabilities like image generation, video creation, Deep Research, and Deep Think.

Key Factors and the New Usage Dashboard

According to Google, three main elements determine how quickly your compute quota drains: prompt complexity, chat history length, and the size or number of uploaded files. To help users keep track, the Gemini web interface is adding a "Usage Limits" dashboard, displaying current usage percentages, reset times, and weekly caps.

The reset mechanism has also changed to a two-tier system similar to Anthropic's Claude: quotas partially recover every 5 hours, combined with a strict weekly limit. This prevents users from spamming requests in a short period and sets a weekly ceiling.

Plan Multipliers and Transition Period Confusion

Under the new rules, subscription tiers grant different compute weights: the Free tier gets base usage; Google AI Plus ($8/mo) gets 2x; Google AI Pro ($20/mo) gets 4x; and the top-tier Google AI Ultra provides significantly more compute power. Once high-tier compute is exhausted, users are temporarily downgraded to the faster, lightweight Flash model rather than being locked out completely.

However, because Google's announcement was brief, many details remain unpublicized. The official support pages still display old text (e.g., "100 messages per day for Pro 3.1"), suggesting a transitional phase or delayed documentation updates.

Adaptation Strategies for Power Users

This change impacts power users who rely on Gemini for extensive coding, massive file uploads, or Deep Research. To adapt, users will need to keep prompts concise, start new chats regularly to wipe out heavy contexts, and manage advanced features wisely to avoid being downgraded to lighter models prematurely.

References & Sources:

  1. Google Support Official Documentation

  2. PCWorld - "Google just made big changes to Gemini usage limits"

  3. Neowin - "Google silently kills unlimited Gemini access as strict new usage limits roll out"

Frequently Asked Questions

How does Gemini's new "Compute Usage" rule affect casual users?
For casual users who only ask short questions, the impact is minimal since simple tasks consume very few resources. However, power users who frequently feed long text, upload large files, or use Deep Research will find their quotas running out much faster than before.
Will I be locked out of Gemini completely when my compute quota runs out?
No. Under the new system, when paid subscribers exhaust their compute allocation within the 5-hour window, they won't be locked out. Instead, the system will automatically downgrade the conversation to a faster, low-cost model (like Gemini Flash) until the quota partially resets.