Google Replaces Gemini’s Daily Prompt Caps With Compute‑Based Usage Limits

The new calculation considers several factors:

• Prompt complexity: harder reasoning or coding tasks consume more compute.
• Features or models used: advanced tools and multimodal capabilities can increase usage cost.
• Conversation length: long running chats gradually use more quota as the context grows.

In practice, this means one prompt may count far more than another depending on how demanding the request is. A simple text question might use minimal quota, while tasks such as research, coding, or media generation may consume significantly more.

Five‑hour refresh windows and weekly ceilings

The new system also introduces a different rhythm for resets.

• Usage refreshes every five hours, giving users additional quota periodically.
• These refreshes continue until a broader weekly limit is reached.

Instead of a single daily reset, users now operate within rolling windows of availability, which can allow bursts of activity but still enforce an overall weekly cap.

According to Google’s support documentation, these changes apply to users aged 18 and older, while younger users were not initially affected by the update.

How subscription tiers fit into the system

Google introduced and reshaped several consumer AI subscription tiers around the same time.

Key tiers reported around the rollout include:

• AI Plus (~$8/month): entry‑level paid tier with higher usage limits than the free plan.
• AI Pro ($20/month): expanded access to models and features with larger usage allowances.
• AI Ultra ($100/month): a newer plan aimed at advanced creators and developers with substantially higher limits.

Reports also noted that Google adjusted the pricing structure of its highest tier, introducing the $100 AI Ultra plan while lowering the cost of the previous top plan.

While exact quotas vary by plan and task type, Google states that paid tiers receive higher compute limits and priority access to models and features.

What happens when you hit your limits

When users exhaust their quota, Gemini may shift interactions to smaller, faster models rather than stopping completely, allowing basic functionality to continue.

Google also supports purchasing additional AI credits for some services or upgrading to a higher subscription tier to increase available usage.

However, details about exact downgrade behavior or the precise mechanics of credit purchases vary by feature and plan.

Why some users are criticizing the change

The rollout generated backlash from some Gemini users, particularly among paid subscribers.

Several complaints center on how the new quota system affects real‑world usage:

• Perceived reduced value: some users argue the new limits effectively reduce usable output despite similar pricing.
• Long chats consume more quota: extended conversations increase compute usage because the model must process the entire context.
• Plan changes: reports indicate the AI Pro plan removed the previously included 1,000 monthly AI credits, which some users viewed as a downgrade.

Critics described the change as a "bait‑and‑switch" because the pricing remained similar while the mechanics of usage became stricter.

A broader shift in AI pricing models

Google’s move reflects a broader trend in AI services toward compute‑ or token‑based pricing instead of simple message counts.

Under this model, platforms measure the actual computational work performed rather than the number of requests.

This approach is increasingly common across advanced AI systems because:

• complex tasks require dramatically more compute than simple ones
• multimodal tools (image, video, coding agents) can be far more resource‑intensive

By tying usage limits to compute, providers can allocate resources more efficiently as AI systems become more powerful and expensive to run.

The bottom line

The May 2026 update marks one of the biggest structural changes to Gemini since launch. Instead of counting prompts, the platform now measures how much work each interaction requires.

For casual users, the experience may feel similar. But for heavy users—especially those running long conversations, complex prompts, or advanced tools—the new compute‑based quotas can cause limits to arrive much sooner than under the previous daily prompt caps.