AI Ultra Plan Quota Limits Killed After Latest Update

Google Antigravity
1

I recently upgraded to the AI Ultra Plan, been using it daily for ~12 hours a day for the last 2-3 weeks. Using Claude Opus almost exclusively and only hitting quota limits 1x a day and easily managing with the 25K extra credits in between the wait periods. Felt like nearly unlimited use for my use case up until 2 days ago after I updated the application.

Was using Claude Opus yesterday, hit the quota limit, then checked my credits. Had over 20K remaining a few days ago, now it’s down to <10K. Went to lunch and waited a few hours, came back and started using it again, hit the quota limit within an hour. Switched to Gemini to complete the work I needed then stopped completely at that point.

Today at the start of a new session, I see that my quota has dropped 20% after 3 prompts using Claude. (Copy file from docker container volume, query SQL DB, edit env var names in docker compose file).

This is now worse than the usage limits when I was paying for Pro 4 weeks ago. I paid over 10x the price of Pro for AI Ultra and expected >10x the usage, which I was getting initially. Now it’s been nerfed to like <10% of what it was 2 days ago.

If this continues I guess I will cancel my subscription and go back to writing code manually like a caveman. If a mega company like Google can’t even work out the economics of running LLMs without screwing their customers over, the future doesn’t look good for any of these AI companies.

2

Plenty of other options out there. I keep ultra for deep think basically. They don’t care about quota issues, you were a cost to them running Opus 12 hr a day, cheaper for them if you stop paying.

Reality of where AI is heading, personally I would get very comfortable working with cheaper/lesser models

3

Yeah, if that’s the case I will save them some money and cancel my Ultra subscription in favor of Claude.

4

@ggarcia209 I’m curious if you think Anthropic will follow a similar trajectory, and if so, which direction you’d lean. As @Joao_Paulo_Leite noted, reality is hitting advanced models harder this year……have you considered DeepSeek or other non-US models as viable alternatives?

5

I think Anthropic certainly will eventually. At this point, running a Claude Pro plan + DeepSeek seems like a much better alternative than what Google is offering.