Critical Performance Issue – Gemini 3.8 Flash Extremely Slow in Antigravity IDE 2.31.2 Despite Google AI Pro Subscription

Dear Google Antigravity Engineering Team,

I am reporting a serious performance issue with Gemini 3.8 Flash in Antigravity IDE.

My environment:

- IDE: Google Antigravity

- Version: 2.31.2

- Model: Gemini 3.8 Flash

- Subscription: Google AI Pro (activated through Jio recharge offer)

- Operating system: Windows 11 ARM64

- Device: Lenovo IdeaPad Slim 3x (Snapdragon X)

- Network: BSNL FTTH high-speed broadband

- Region: India

Problems observed:

1. Even simple prompts take an unusually long time to receive a response.

2. Coding tasks remain in the “Working…” state for extended periods, sometimes exceeding one hour.

3. Gemini 3.8 Flash has unusually high response latency.

4. Other models available in the same Antigravity installation respond normally and much faster.

5. My AI usage limit is not exhausted, and my internet connection is functioning normally.

Troubleshooting performed:

- Tested simple prompts in fresh conversations.

- Compared Gemini Flash with other available models.

- Confirmed that other models work normally.

- Restarted and tested Antigravity.

My concerns:

I have an active Google AI Pro subscription, but Gemini 3.8 Flash performance is significantly below expectations.

Why does the model take so long when other models work normally in the same environment?

Is this caused by backend capacity exhaustion, regional routing, account-specific resource allocation, or a model-serving issue?

Why does Antigravity remain stuck on “Working…” instead of displaying an appropriate server error when requests cannot be processed?

Requested action:

1. Investigate Gemini 3.8 Flash serving capacity and regional routing for Indian users.

2. Check whether Google AI Pro users are experiencing model-specific backend allocation problems.

3. Investigate repeated retries and excessive response latency.

4. Provide a clear explanation and estimated resolution timeline.

5. Improve error reporting instead of leaving users indefinitely in the “Working…” state.

I would appreciate an engineering team’s response and a permanent resolution.

Thank you.

me to it takes f.. hours to modify a simple css edit !!! i have 2 accounts paid pro and what the hell the are taking hoursssss to add or edit a simple task , they even get more stupid

same here. 3.8 flash on high is insanely slow since yesterday

Same here, it was ok when it started but in last couple of weeks when argon released the 3.8 became supper slow!!!

Same! Before I thought it’s about some errors in updating tokens when switching my accounts, soft issues, etc.
But now I’ve completely erased antigravity and it’s files (except conversations), reinstalled it cleanly.
Still “2+2” takes two-three minutes to answer.
even on my second laptop

Same here, even a simple task takes like mins to respond. Feels really frustrating

Same here!!!
simple task get 3hr time.

Just bring back older, faster models google… 3.7 flash was old but gold standard for most day to day tasks… 3.8 from start till now has been filled with issues, whether token efficiency drops or now becoming less intelligent and much much slower or downright dead. Gemini 4 models are too powerful for simple tasks and consume much more tokens. Bring back 3.7 or fix 3.8 please.

I’m seeing the same thing on the Jio Pro plan.

Latency: Responses are extremely slow, and time to first token is very high, often long enough that the session feels hung.

4 Hydrogen: Locked on my account despite the Pro plan.

3.8 Flash (High): Effectively unusable. Requests appear to deadlock and never complete.

This makes the plan close to unusable right now. Could someone from the team confirm whether this is a known issue with Jio Pro accounts, and whether 4 Hydrogen access is expected for this tier?

I also have the same issue right now. If they’re going to take the 3.8 flash away and replace with 3.7 flash, I’ll be okay with that at this point. what is this simple task working for about 2hrs now and still no output.

This dude @Rebbanpalli_Naveen claimed that 4 Hydrogen available on all plans, including free tiers. And this site Choose a model | Google Antigravity Docs confirms it.

However, mine is greyed out. And they just keep telling to “update the Antigravity” yada yada yada while I’m already on the latest version.

Yep, and it goes even further. The Choose a model page has a picker mockup for each tier, and Gemini 4 Hydrogen is listed as available on Free & Plus (it’s actually the default there) and on Google AI Pro (Trial) with no dagger. Only Argon and Orchestrator have the † “paid Pro or eligible premium device” note.

So by Google’s own docs, Jio Pro should at minimum get Hydrogen. Instead it shows as locked, which means Jio Pro is currently worse than a free plan in Antigravity.

The same issue I’m facing with Google Antigravity. I think this is because of Google itself, They don’t want to give any benefits to the JIO X Google AI pro users.

Loading…Loading…Loading… I’m tired!

Here the same… That are turning impossible to use. There is a reason for that ? I’m awaiting a lot for simple requests.

3.8 is a pile of hot trash. Since 3.1 pro went away I had to switch to Claude.

Too Slow , gemini 3.8 flash . from. JIO user

(post deleted by author)

@frank_zhang Exactly — and it gets even worse. Because Jio Google AI Pro subscribers are locked out of Gemini 4 Hydrogen and Argon, we are forced to use Gemini 3.8 Flash in Antigravity, which is not only painfully slow, it even hallucinates that current 2026 models like Gemini 4 Argon, Claude Sonnet 5.5, and GPT-6 Astra don’t exist (see attached screenshot from my Antigravity session).

Google’s own Antigravity documentation (antigravity.google/docs/models/) explicitly lists Gemini 4 Hydrogen as available across all plans (including Free, Plus, and Google AI Pro), yet Jio-bundled Pro accounts still have it marked as "Unavailable" on the server side.

With legacy models retiring tomorrow (November 18), Google needs to fix the server-side Jio partner SKU entitlement mapping immediately so we aren’t stuck on a broken, slow Gemini 3.8 Flash.

Same here. A small fix in a skill has already taken almost 30 minutes, and it’s still not done!