Google AI Ultra Max 20x Quota Issue: 20% Weekly Limit Gone in One Hour

[moonair]

Public warning for Google AI Ultra Max 20x / Antigravity users.

I was forced to migrate from Workspace Antigravity to Google One / Google AI Ultra.

Since that migration, Antigravity has become practically unusable for my professional development work.

I am subscribed to Google AI Ultra Max 20x, the highest plan available. However, with the exact same workflow I had before the migration, Antigravity consumed around 20% of my weekly Claude/GPT quota in about one hour.

This never happened before the forced migration.

The most concerning part is the lack of transparency:

• No clear details about the quota before the migration.
• No clear details about the quota after the migration.
• No daily usage breakdown.
• No model-by-model consumption breakdown.
• No clear explanation of why the same workflow now consumes so much more quota.

Google Support told me that my account is synced correctly and that I simply reached my usage limit. They suggested enabling AI credit overages, meaning I would need to pay extra credits on top of the highest subscription plan.

That is not an acceptable answer.

The issue is not that I reached a limit. The issue is that the included Ultra Max 20x quota is now consumed extremely fast after a forced migration, making the product unusable for serious professional work unless I pay extra.

If this is expected behavior, then the new Google AI Ultra Max 20x / Antigravity plan is a major downgrade compared to the previous Workspace Antigravity experience.

If it is a bug, Google needs to acknowledge it, investigate it, and fix it.

I am sharing this publicly because I am not the only user reporting similar quota issues today. Developers and professional users need transparency before relying on Antigravity for real work.

Google should provide:

1. Previous Workspace Antigravity quota details.
2. Current Google AI Ultra Max 20x quota details.
3. Daily usage breakdown.
4. Model-by-model consumption breakdown.
5. A clear explanation of the quota changes after migration.
6. A fix or compensation for affected users.

Right now, paying for the highest tier and being unable to work after around one hour is unacceptable.

[Dawid_M]

As I analyze this case, this is not a bug or a problem, these are the current limits. I have regular pro and sometimes I’m tempted to do something, but I have my limit exhausted in 4 to 5 minutes. So your Ultra 20x 20% probably falls exactly within that range you pointed out.

[Faheem_Anis1]

What did you do in that one hour? If you were running an agent continuously without stopping for a whole hour, this is about 1.4m tokens (generally the 5-hour limit for ultra).

If you were running 5 agents concurrently, this is enough to exhaust the quota in one hour, and consume about 50% of the weekly quota.

[moonair]

Your comment misses the point.

I’m comparing the exact same usage before and after the migration from Google Workspace AI Ultra to Google One AI Ultra.

Before the migration, I could use Claude for the full 5-hour window without ever hitting the limit.

After the migration, with the exact same workflow, one single agent, and small, targeted prompts, I now hit the limit after a fraction of that usage.

So this isn’t about running multiple agents or consuming millions of tokens. The only variable that changed is the account migration.

At this point, the most likely explanation is either a bug affecting migrated accounts or a deliberate change in quota allocation after the migration.

What makes this even more suspicious is that Gemini models remain almost untouched. My Gemini weekly quota is still around 98%, and even after hours of usage I never come close to exhausting its 5-hour limit. Meanwhile, Claude and GPT quotas are disappearing at an absurd rate under the exact same working conditions.

If this behavior is intentional to encourage users to switch to Gemini models, Google should be transparent about it. If it’s a bug, it needs to be fixed. Either way, the current experience is unacceptable for users who subscribed to AI Ultra expecting the same level of service they had before the migration.

[DrQwertySilence]

This happens because the model quota factor only applies to Google Models quota, not to the Other models quota (third-party models). That’s why your Gemini quota doesn’t move at all, while your Other models quota does. It was mentioned in a post months ago when they introduced the USD $100 Ultra plan and the Gemini model quota merge. So it is working as intended.

Cheers!

[Faheem_Anis1]

As expected, the non-Gemini quota is much lower than other models. Once you migrated, Google applied the baseline quota provided to everyone else.

If you hadn’t migrated, the quota might have remained for 1-2 months due to subscription terms. I recommend migrating to Gemini 3.5 flash anyway, its as strong as Sonnet and you will get sufficient quota to fit your daily needs

Do let me know if I’m unable to explain this.

[theflowingsword]

I have exact same issue. Before i could run 8 agents, all runnning claude opus 4.6 for 2.5 hrs and complete 15prs of reasonable complexity and all the planning stages in a 5hr window. Now on personal account I get 3 agents using the 5hr limit in 45mins. I will consume the weekly quota for claude in 5 sessions leaving me nothing for rest of week. On enterprise I could complete 60prs on a weekend which is 2 max sessions Friday and 3 on Sat and Sunday. Now i can’t get close to this. I can’t use antigravity with these limits. I did 7 months of antigravity. I can demonstrate my output before and after the move from enterprise. I am now looking at cursor as a new platform. I have tried to talk to support but all I get is “its an automated limit, its calculated correctly” ie computer says no. not much use if you are a coder who has deadlines and projects to deliver. very dissappointed