Using Antigravity today has been absolutely terrible! Slow, constant errors, models doing really stupid things (recursive nonsense and general bad stuff) and I am positive because of these issues my quota has gone down significantly. I can see multiple reports of this, clearly something has gone wrong today.
I am just about to run out of my weekly limit and today has been a big factor to this. When Google has issues like this it would be really useful to have some quota reimbursed or reset, even if it were back to the start of the day.
Thank you for reporting this. The slowness and unexpected model behavior you experienced were the result of a temporary backend service disruption.
While manual quota resets or individual token refunds cannot be issued, model quotas in Antigravity operate on rolling time-window allocations that automatically replenish as time passes.
To restore optimal performance and work around high-demand periods:
Verify current service connectivity by opening a new chat session.
Switch to Gemini Flash models (e.g., Gemini 3.7 or 3.8 Flash) in the model selector, as Flash models utilize separate, higher capacity allocations with significantly higher throughput.
If your session or authentication state feels unresponsive following the incident, open the Command Palette (Ctrl+Shift+P on Windows/Linux or Cmd+Shift+P on macOS) and run Antigravity: Reset Onboarding to refresh your local token cache.
Start a ‘New Chat’ session for new tasks to avoid carrying bloated conversation history from earlier failed turns
I have tried all of that, but the issue persists. The model keeps freezing in place, and I have to manually stop it and say “continue” for it to generate a bit before halting again.
This is only happening with one specific account; my other account is working completely fine, and I don’t know why both are on the Pro plan.
Additionally, I couldn’t run the Antigravity: Reset Onboarding command. When I search for it in the Command Palette, it doesn’t appear at all.
Same, last friday the 3.8 high was perfect, it completes a 2+ hour task without significant errors.
Today same configuration with a similar task is running 4+ hours (it seems “more stupid” and “more slow”).
yea ive noticed this too since yesterday. it’s been excruciatingly slow and i frequently get errors, prompting me to retry. im on the highest plan too so it feels like kind of a waste not getting everything out of my weekly limit
Since it’s built on Electron, it will always run like utter garbage on Linux. I have to restart it in my terminal after every 2 chat turns because it freezes. Luckily I got unlimited Anthropic credits for the past 2 months so I ditched it and have been usijng VSCode again with a fork of Codex that I customized for my own purposes as the harness. That and Google got so stingy and tight with the credits, and the models got SO unwilling to work on anything related to cybersecurity (despite being… at it compared to Opus and Fable) that it’s like I just use Gemini now to answer stupid stuff like “hey, Gemini, why isn’t this screw drilling into the wood beam properly? Do I need full threaded or part threaded?”
I have the same issue, it is so slow, yesterday it was not. Did google decide to cut resources for some locations? How can we see antigravity performance accross the locations.
