[Bug] Persistent "Model unavailable" loop on Gemini 3.7 Flash High (Pro Account)

Issue Overview: I am encountering a persistent connectivity error while using the Gemini 3.7 Flash High agent on Antigravity. I am an active Pro Plan subscriber.

Error Message: "Model unavailable, retrying in 3s (attempt 2/9), Agent execution terminated due to error."

Steps to reproduce / Behavior:

  1. Select the Gemini 3.7 Flash High agent/model.

  2. Submit any prompt.

  3. The request hangs and enters an infinite retry loop without ever completing.

  4. Workaround: Changing the agent/model to a different one immediately resolves the issue and prompts are processed normally.

Environment Details:

  • Account Tier: Pro Plan

  • Affected Model/Agent: Gemini 3.7 Flash High

It seems like the request pool or API route for this specific high-reasoning agent is hitting a bottleneck or server-side failure.

I have attached a screenshot showing the error interface. Any insights or fixes from the team would be greatly appreciated!

Hi @bito_bobli,

Thanks for the screenshots! “Model unavailable” means Gemini 3.7 Flash High couldn’t respond at that moment. That’s on our side, not your account, which is why switching models works.

I also noticed this conversation is quite long (2 days, with many steps), and you’re asking it to continue. Very long conversations can sometimes keep failing on one model. Please try:

  1. Start a new conversation with Gemini 3.7 Flash High, give a short summary of where you left off, and continue from there.
  2. If it still fails, use another model for now and try High again later.
  3. Update Antigravity to the latest version.

If the new conversation hits the same error too, let us know!.

“your broken system. Stop wasting my time with these endless retries and get your together!”

Are you kidding me?! What kind of system are you running? Every single time I try to use your service, I get the same: “Model unavailable, retrying in 26s.” Then I wait, only to get another error. What the actual is wrong with your servers?!

I’m not here to waste my entire day staring at a loading screen and watching your broken system struggle to do the one thing it’s supposed to do. If your models can’t handle basic requests, then why the are you even offering this service? Stop pretending everything is working when your infrastructure is clearly falling apart.

Your endless retries are not a solution. They’re just another excuse for a system that can’t do its job properly. I don’t give about your countdown, your automated messages, or your useless error notifications. I want a service that actually works!

Fix your unstable servers, fix your model availability, and stop treating users like unpaid beta testers. People are paying for a service, not for the privilege of watching your incompetent system crash, retry, and fail over and over again.

Get your act together. This is beyond frustrating, completely unacceptable, and honestly one of the most embarrassing user experiences imaginable. ENOUGH WITH THE RETRIES. MAKE THE MODEL AVAILABLE AND DO YOUR JOB!