Persistent 503 MODEL_CAPACITY_EXHAUSTED error across all models (Antigravity / Gemini / Claude)

Google Antigravity
1

Hello Google Support Team, I am experiencing a persistent issue while using the Antigravity / AI coding environment. Problem: I consistently receive the following error across multiple models: HTTP 503 Service Unavailable MODEL_CAPACITY_EXHAUSTED “No capacity available for model on the server” This happens even when switching between different models. Models affected: gemini-pro-agent claude-sonnet-4-6 other available models (all tested) What I have already tried: Switching between all available models Creating and testing with different accounts Using VPN (ProtonVPN, United States / Miami server) Restarting sessions and creating new sessions Completely reinstalling the Antigravity application Clearing cache and local data Reducing prompt size and simplifying requests Main use case: I am running a workflow where the system analyzes and prepares previously generated project files / codebases. The prompt typically involves: “analyze and prepare existing project structure / files” Even simplified prompts fail with the same error. Result: No configuration or workaround has resolved the issue. The system either: fails immediately with 503, or gets stuck in “working/loading” state without producing output. Request: Please check whether: there is a regional capacity limitation my account is rate-limited or flagged the models are currently under heavy load or restricted for my environment Antigravity service has known outages or backend restrictions I would appreciate any guidance or fix, as the issue blocks all usage completely. Thank you.

2

This is also happening to me for literally 13 days already, and there’s no solution for that.

3

I am experiencing the exact same issue for the last few days. Despite being on a Pro/Ultra plan, all models return the 503 MODEL_CAPACITY_EXHAUSTED error immediately.

I’ve noticed a few things that might be relevant for the Support Team:

  1. Backend vs. Local: Like Karen, I’ve tried clean reinstalls and switching accounts/VPNs, but the issue persists, which strongly suggests a backend routing problem rather than a local setup issue.

  2. Possible Endpoint Bug: There have been reports (specifically regarding WSL/Remote environments) where Antigravity incorrectly hits the daily-cloudcode-pa.googleapis.com (staging) endpoint instead of production. Even if not using WSL, it feels like certain accounts are being routed to low-capacity nodes.

  3. Quota Sync Issue: In some cases, the quota is being depleted even when the model fails to respond with a 503 error.

To the Google Team: Could you please verify if there is a known service disruption for the cloudcode-pa endpoint or if there’s a regional capacity rebalancing happening? This is a complete blocker for professional workflows.