Google Gemini's new credit system: how it works, what you get

Key Takeaways

• Gemini now charges by compute intensity, not request count. Complex prompts burn credits faster.
• Four paid tiers: AI Plus ($8/mo), AI Pro ($20/mo), AI Ultra ($100 or $200/mo) with 2x, 4x, and 5-20x the free limits.
• Usage resets on two cycles: every five hours for burst limits, every week for the cap.

Google has overhauled how it meters Gemini AI. Instead of counting requests, the system now tracks compute intensity. A simple weather query costs less than a code-generation prompt or a video render. The change makes billing more efficient for Google but leaves users guessing when they'll hit their ceiling.

Why Google switched to compute-based credits

The old model was blunt: X image generations per day, Y prompts per hour. That ignored the reality that generating a two-minute video eats far more GPU cycles than summarizing an email. Under the new system, each request drains a variable amount of credit based on the model you select, the thinking depth you choose, and the length and complexity of your prompt.

Google's support docs are candid about the tradeoff. Limits are 'subject to change or may be limited based on testing, experimentation or availability.' Translation: on busy days, you might hit walls earlier. Free users sit first in line for throttling.

The four paid tiers and what they include

Google offers three subscriptions in the US, stacked on top of a free tier:

• AI Plus ($8/month): 2× the free limits.
• AI Pro ($20/month): 4× the free limits.
• AI Ultra ($100 or $200/month): 5× or 20× the AI Pro limits, depending on which Ultra plan you pick.

All tiers unlock every Gemini model: Flash-Lite, Flash, and Pro. Higher models are smarter but drain credits faster. Each model also offers three thinking modes: Standard, Extended, and Deep Think. Extended and Deep Think produce more nuanced answers at the cost of speed and usage budget.

Context windows by tier

Context window size determines how much text Gemini can hold in memory during a single conversation. Here's the breakdown:

For builders shipping agentic workflows or long-document analysis, that jump from 32K to 1M tokens is the headline feature. It's what separates a toy demo from a production system.

How to check your Gemini usage

Google at least makes monitoring easy. On the web app, click the cog icon in the lower left and select Usage limits. On Android or iOS, tap the menu button, then the cog, then Usage limits.

You'll see two progress bars. The first tracks your rolling five-hour quota. If you burn through it, you wait until the next reset. The second bar shows your weekly cap. Exhaust that on a paid plan and you're demoted to the most basic model until the week rolls over.

What this means for teams building on Gemini

If you're prototyping on the free tier and suddenly hitting walls, the new credit math is probably the culprit. A workflow that ran fine last month might now stall because it involves heavier prompts or more video calls. Budget accordingly: for any production use, AI Pro or Ultra is the realistic floor.

The opacity is the real problem. Google publishes multipliers but not absolute numbers. You can't plan a sprint when you don't know your actual capacity. Competitors like OpenAI publish explicit rate limits for API tiers. Google's consumer-first framing doesn't translate well to engineering estimates.

Frequently Asked Questions

How does Google Gemini's new credit system work?

Gemini now measures usage by compute intensity, not request count. Complex prompts, video generation, and advanced models consume more credits than simple text queries.

What are the Gemini subscription tiers and prices?

Google offers AI Plus at $8/month, AI Pro at $20/month, and AI Ultra at $100 or $200/month. Each tier multiplies the free tier's limits by 2×, 4×, or 5-20× respectively.

How often do Gemini usage limits reset?

There are two resets: a burst limit that refreshes every five hours and a weekly cap. Both are visible in the Usage limits section of the Gemini app.

What happens when I hit my Gemini usage limit?

Paid users get demoted to the most basic AI model until the next reset. Free users may face longer waits or stricter throttling during high-demand periods.

What is Gemini's context window size by tier?

Free users get 32K tokens (~24,000 words). AI Plus gets 128K tokens. AI Pro and Ultra users get 1 million tokens, roughly 750,000 words per conversation thread.

Source: Feed: Artificial Intelligence Latest / David Nield