I am experiencing something extremely similar. It is highly disappointing I couldn’t code since the morning because anti-gravity and gemini cli are not responsive. Notice I am a Google AI ultra user. I should be getting top service and I switched from Claude code to try out Google’s coding ecosystem for AI, I highly un-recommend it. Maybe I should switch back. Extremely disappointing, Google, and I reached out to you guys. You guys didn’t update me on why anti-gravity is not working and gemini cli is not working. What is wrong with your products? Your competitors never do this.
Can I code in peace without getting hit with rate limits? What is wrong with you google?
When purchasing the maximum subscription, you claim that users receive full access to the models. In reality, not only is that access unavailable, but even the access that does exist performs significantly worse according to benchmarks (including due to frequent failures).
This means that if we are paying $250 and 7 out of 10 requests result in errors, you should compensate for the time and resources wasted due to your fault. It is impossible to work under these conditions. The failures are constant.
If you are unable to provide a stable and reliable solution, why release it to the market at all?
I am posting this as a paying customer on the AI Ultra Access add-on for Google Workspace (the highest-tier enterprise plan).
We subscribed specifically to use Antigravity for serious agentic development work. However, we are constantly hitting MODEL_CAPACITY_EXHAUSTED errors (HTTP 503 Service Unavailable) within minutes of starting any meaningful session — especially with Claude Sonnet 4.6 and Gemini 3.1 Pro models.
The errors happen so frequently that the platform is effectively unusable for any real agentic workflow. This is despite the “generous quotas” and priority access that were advertised for the Ultra Business tier.
This is not an occasional issue — it has become a daily blocker for our team. As an enterprise customer on the top plan, this level of capacity exhaustion is simply not acceptable.
We have already reported this multiple times through Workspace support and in-app feedback, but the problem continues.
Questions for the Google / Antigravity team:
Is there a known capacity issue affecting Workspace Ultra users right now?
When can we expect a meaningful increase in available capacity for the Ultra Business tier?
Are there any recommended workarounds or model settings that actually help reduce these errors?
Would love to hear from other Ultra Business / Workspace users — are you facing the same problem?
@GoogleStaff / @Antigravity-Team — any visibility or ETA on a fix would be greatly appreciated. Happy to provide more details privately if needed.
11 crashes for one text window request. My subscription was renewed for a third month yesterday. It’s the maximum plan.
My colleague, who works outside of Antigravity, on Claude Code, doesn’t have these crashes. For you, using Claude Opus, these crashes are CONSTANT.
{
“error”: {
“code”: 503,
“details”: [
{
“@type”: “type.googleapis.com/google.rpc.ErrorInfo”,
“domain”: “cloudcode-pa.googleapis.com”,
“metadata”: {
“model”: “claude-opus-4-6-thinking”
},
“reason”: “MODEL_CAPACITY_EXHAUSTED”
}
],
“message”: “No capacity available for model claude-opus-4-6-thinking on the server”,
“status”: “UNAVAILABLE”
}
}
This is ridiculous it’s now falling down the rabbit hole trying to fix the errors the previous crash caused. getting worse and worse until it restores from git losing all the work it had done to that point.
