Gemini 4 Argon (High) fails on every prompt in Antigravity IDE, CLI and app: "Model unavailable, retrying (attempt n/9)" then "Agent execution terminated due to error" (Google AI Pro)

Google Antigravity
1

SUMMARY I use Antigravity from Bangladesh with a personal Google account that has an active Google AI Pro plan. With Gemini 4 Argon (High) selected, every prompt fails with repeated “Model unavailable, retrying in Ns (attempt n/9)” messages and finally “Agent execution terminated due to error.” The same problem happens in all three places: Antigravity IDE, Antigravity CLI (agy) and the Antigravity app. Everything is fully updated, and the Usage panel shows quota remaining, so this does not look like a quota problem. VERSIONS - Antigravity IDE: 2.6.0 (update check says no updates available) - Antigravity CLI: 1.5.4 - VSCode OSS: 1.107.0 - OS: Windows 10.0.26300 (x64) WHAT HAPPENS 1. I send any prompt with Gemini 4 Argon (High) selected. 2. The agent starts normally and runs its first steps (environment check, media probe, creating a script, audio analysis). 3. Then the status shows “Model unavailable, retrying in 8s (attempt 6/9)”, later “retrying in 31s (attempt 7/9)”. 4. After about 9 minutes it ends with “Agent execution terminated due to error.” Clicking Retry does not fix it. USAGE PANEL (same session) - Gemini Models: weekly limit 100% left, 5h limit 100% left - Claude and GPT models: weekly limit 100% left, 5h limit 100% left - Gemini 4 Argon: weekly limit 92% left (resets in 6d 17h) - Gemini 4 Carbon is listed as Unavailable. EXPECTED Prompts run normally while quota is available. ACTUAL Prompts fail after repeated “Model unavailable” retries in the IDE, CLI and app. REQUEST Please check whether Gemini 4 Argon has an availability or capacity problem for Google AI Pro accounts, or whether my account entitlement needs a reset.

Screenshot 2026-11-11 173728
Screenshot 2026-11-11 173728806×851 33.3 KB

2

Are you having normal pro plan or Paid Pro plan?

3

This happens also with Ultra lately including with 3.8 Flash, 4.0 Argon or 4.0 Carbon

A temporary solution is too long out and log in again with a different account, whig I believe lands on a different server.

I think the error is trying to say servers are too busy, cool off for a period of time

4

Hi @Jubayir and @softer,

Welcome to the Forum,

Thank you for the detailed report and screenshot. Here is what is happening with your Gemini 4 Argon (High) requests and model picker:

  • 1. Entitlement vs. Server Capacity (Model unavailable, retrying): Because your Usage panel shows 92% weekly Gemini 4 Argon quota remaining, your Google AI Pro entitlement is active and does not need an account reset. The repeated "Model unavailable, retrying" errors after the initial tool steps are caused by temporary regional server capacity spikes (HTTP 503) on Gemini 4 Argon (High), which our engineering team is actively addressing.
  • 2. Why Gemini 4 Carbon Shows as Unavailable on Pro: Per our Models and Plans documentation, Gemini 4 Carbon requires a Google AI Ultra (or Gemini Enterprise) plan, so it is expected to appear under Unavailable on Google AI Pro.
  • 3. Recommended Workarounds: Please update the Antigravity CLI from v1.5.4 to v1.5.5 (agy update, Changelog), switch the thinking level to Gemini 4 Argon (Medium) or Low (or Claude Sonnet 5.5 / Gemini 3.8 Flash) during peak load, and start a fresh conversation if media/audio probing steps have created a large context history.