Posting this again because it hasn’t been fixed and it’s getting worse.
Ultra subscriber. Agent terminated, 503, no capacity. Over and over. I can get maybe a few requests through before it locks up for the rest of the session. I have 5 hours of daily quota and I burn through it in under an hour because every terminated attempt still eats into it. Most of that time isn’t productive work. It’s retries.
The Workspace account situation is especially bad. I test on both a personal and a Workspace-linked Ultra account. The Workspace one is significantly worse, fails faster and more often. Free accounts don’t work at all anymore. Instant termination. Not “sometimes fails,” not “intermittent.” Just dead.
Here’s the thing that keeps bugging me. Normal software has users complaining about features, about how things work. Antigravity has users complaining about not being able to use it. That’s a different category of problem entirely. The #1 issue right now isn’t model quality or missing features. It’s that the product doesn’t turn on.
I’ve posted about this before. Others have posted about this. The threads are all over this forum. Whatever capacity provisioning or routing is happening on the backend, it’s broken, and it’s been broken for weeks.
3 Likes
You’re not alone in this!
1 Like
your account are shadow banned (or some mechanism that target account)
try to switch to another account if you have one, my other account work almost perfectly.
1 Like
I had an update today, and it seems to have fixed the issue. Knock on wood…I haven’t had to click retry once today. I’m on Antigravity Version 1.23.2 after the update this morning.
1 Like
so is there a way to get around this?
It’s working for me today too but I don’t think it’s the update. This update dropped like a week ago, and the pattern I’ve been seeing is time-based. It usually starts working around 1-2 AM PKT and holds until about 5-6 AM, then dies again. Today it’s lasted longer than usual but that’s happened before.
I’m actually on one version older than 1.23.2 because the latest introduced a bug where you get a “model has exceeded the output token” error mid-response. Rolled back and that went away. Still getting the same capacity pattern on both versions though, so I don’t think the update changed anything on the 503 side.
Hoping I’m wrong and it actually stays up this time.
I have one computer that started giving 503’s after upgrade to 1.23.2… Not sure if it’s related, but this is when it started. I didn’t succeed in any AI calls after the upgrade and stopped trying. Luckily this was not my main computer, that one is still on 1.22.2 and never had the 503 issue. Same account. Won’t upgrade that one till the other one is working again, not sure if this will require the next update.