Antigravity is in preview and active development. Even though they accept payment for it, it is in no way production grade. And given the quality of their releases, the stability of their infrastructure, and the complete absence of support, I definitely would not feel comfortable saying Antigravity has “enterprise-level security and data protection”. There is nothing about this project that says “enterprise-level”.
1 Like
I have to strongly disagree with you in terms of provided security guarantees. You are conflating the free individual preview with the paid business tiers.
We are paying (I do) for the Google Workspace AI Ultra for Business plan. Under Google’s own documentation, team usage is governed by Section 5 of the General Service Terms in Google Cloud’s enterprise terms of service. That legally guarantees a baseline of enterprise-level data protection and IP security.
A piece of software can be in technical ‘preview’ while still being wrapped in an ironclad, enterprise-grade security contract. Google is legally guaranteeing zero data harvesting of our code and binding the tool to their Cloud Enterprise Terms of Service (I am talking about Google Workspace AI Ultra for Business plan here). They are charging us top-tier prices for that exact IP protection and enterprise compliance. Therefore, expecting the infrastructure to actually be stable enough to use those secure features is completely justified.
2 Likes
Yeah, I’m completely on the same page with this.
I actually hadn’t seen 503 errors for a while, but lately, they’ve been popping up way too often.
Sure, hitting “retry” or asking the model to “continue” works as a temporary band-aid, but it’s really not a proper fix. Having the generation interrupted and forcing a “continue” mid-way totally breaks the workflow and definitely messes with the quality and consistency of the final output.
Right now, it’s getting to the point where I’m hitting a 503 almost every minute even when I retry, which makes getting any real work done practically impossible.
I strongly agree that there needs to be some kind of priority system. If we’re paying for the Enterprise/Ultra tier, shouldn’t we be getting guaranteed capacity or priority routing when the servers get slammed?
1 Like
Absolutely there is no peak when it 24/24 I did try to use it every hour to see and there is always overload any hours day,night or early morning..
1 Like
I have never had this on Ultra. I am using Opus 4.6 without any issue right now. How often does it happen to you?
Shouldn’t Ultra users get priority during peak hours? The constant 503 errors are making it completely impossible to get any work done right now.
1 Like
This is very frustrating, it’s been going on for 2 months already. Until yesterday I was using antigravity version 1.19.6 because this problem didn’t occur in that older version, but now Google has deactivated that version and it’s unbearable.
And it’s not a matter of maximum capacity because I log out of my Google Workplace professional account and log into my personal account and mysteriously there is no more “capacity exhausted” message.
yes I am completely on the same page with this….
I am beyond frustrated. I’ve been paying for the Ultra Tier specifically to avoid service interruptions, yet for the past 48 hours, my workflow in Antigravity has been completely paralyzed.
I keep getting hit with HTTP 503 Service Unavailable and MODEL_CAPACITY_EXHAUSTED for claude-sonnet-4-6. My logs show the server (ESF) is dropping requests after nearly a minute of “working” time.
Technical details of the latest failure:
-
Trajectory ID: 73ccf327-79a3-4fa8-b4f2-a3b4d517
-
TraceID: 0x5044f7139b2
-
Error: No capacity available for model claude-sonnet-4-6 on the server.
Why am I paying premium prices for a service that treats me like a free-tier user during peak hours? This downtime is costing me significant time and money. At roughly $10/day for this subscription, I expect a stable environment, not a “High Traffic” lottery.
Consider this my final notice: this is the last month I am paying for this subscription unless Google fixes the capacity issues immediately. Furthermore, I believe a compensation/refund for these days of total service failure is the bare minimum a tech giant should provide to its professional users.
Fix your infrastructure.
Yea, I was looking forward to my refresh - Instead of a refresh I got ZERO additional token usage and a flat out 20hr wait.
Thanks, I opened Dev to write the same freaking thing. I can’t determine what we’re paying for? Drive storage? AI features? Priority support? None are in a usable condition. Oh god, not even two of the models are working without getting ‘Our servers are experiencing high traffic right now, please try again in a minute.’
I’m getting this too periodically. More in the EU morning - I would have expected it in the US timezone more.
I think what is happening is that the adoption of AI is happening faster than any of the providers can create compute capacity, hence the driver for data centers. The problem is there’s not enough people paying for it yet to actually make it financially viable. The best choice in the interim is this particular error: it’s to limit users’ ability to get things done so that everybody gets a bit of CPU but not all of it.
I kept hitting retry over and over, the task never even ran, and somehow 40% of my tokens were still consumed. How is that even possible?
I see the same issue with all accounts 2 Pro, 2 Ultra!
I also have same issue! Even the model quotation of Ultra plan is full (not yet used today)!
Most likely a problem from Anthropics side. They did not want to invest in their own compute and are now facing severe capacity constraints.