PSA: The "7-Day Lockout" isn't a bug. Here's how the new Hybrid Quota System works.

Google Antigravity
, , ,
1

Hey everyone,

I’ve seen a lot of panic here about users getting hit with massive 4 to 7-day cooldowns on Pro/Ultra models, making them feel like they’re on a free tier. I did some digging, and it’s not a ban or a bug. Google silently switched us to a Hybrid Quota System.

Here is what is actually happening and how to fix your workflow:

1. The “Lockout” is just the new Base Quota

The regular 5-hour refresh is gone for heavy models. Your base quota now lasts much longer, but takes days to refresh.

2. The Solution: AI Credit Overages

As Pro/Ultra subscribers, we now get a monthly stash of AI Credits (1,000 for Pro, 25,000 for Ultra). They don’t roll over. To bypass the long cooldown, you need to enable this feature in the software.

Here is how to find it:

  • Click the Gear icon (Settings) in Google Antigravity.

  • Go to Antigravity User Settings > Models.

  • Right at the top, you will see your remaining credit balance based on your subscription tier.

  • Toggle ON “Enable AI Credit Overages”.

When your base quota hits zero, it will automatically start deducting from your monthly AI Credits instead of locking you out. I have verified this in my Google One credit history: standard coding prompts cost about 3 to 4 credits per request.

:warning: CRITICAL BUG WARNING (HTTP 400)

Right now, if you have Overages enabled and try to attach an image/screenshot in your prompt, the server might reject it with an HTTP 400 Bad Request (Unable to process input image).

Workaround: Stick to text-only prompts (copy-paste your code/errors) while using Overages until Google fixes the vision endpoint.

Hope this helps you save your momentum and project context!

2 Likes
2

Hi @Samber_gg

Thank you for your valuable feedback. We sincerely appreciate you taking the time to share your thoughts with the community. Your input is crucial for our continuous improvement and we have shared it directly with the relevant internal team.

3

Unfortunately, it’s definitely still a bug. I used Claude for maybe 10-12 prompts total. I wiped the context completely fresh every single time, had zero background loops running, and used a strict ignore file to block excess data.

Somehow, the tool still burned my quota from 100% down to 20% on just those 10 prompts and slapped me with a 7-day lockout. To make matters worse, I literally have 1,000 credits sitting in my account, but the system is hard-locking me from sending a single message. Burning that much quota on a handful of zero-context prompts just doesn’t add up.