Antigravity Models are not loading, even tough my account is successfully logged in

[yonnie]

All my quota gone without nothing done because i kept gettin errors and the try again button.
the whole conversation is Continue promt.
quota gone, what a waste of time, i should have not switched from cursor.

[Cypher_One]

Gemini is working fine for me Claude Opus 4.5 is a bit buggy rn in Planning and Fast modes.

[Abhijit_Pramanik]

We appreciate you sharing your concern about this issue and for taking the time to provide further details.

This matter has been escalated internally for a comprehensive investigation. Your input is invaluable as we work to resolve it.

[Segarra]

Same problem for me, once a try to log again, I have this error, and it’s impossible to log with that account.

[ment]

25h !? reary!?

The Pro Plan should be every five hours, but…

[raphaelp]

Same here with Gemini models:

“Agent executive terminated due to error” with “Loading…/Working…/Generating…” message in a loop for 4 days now.

[raphaelp]

Hi, We’re still have an issue with

“Agent executive terminated due to error” with “Loading…/Working…/Generating…” message in a loop for 4 days now.

[wang_daniel]

Model unable to use again,

[ping_blue]

still

[Steve_Bula]

Antigravity must have been developed be Gemini: please fix the bug –> done! (deleted unit tests…)

[Felix_Hartono]

Yes, i think they decreate the rate limit?

[Kef_Iras]

Still not working for me…

[kjsymatio]

2026-01-23 09:23:05.688 [info] W0123 09:23:05.688173  4155 log_context.go:106] Failed to refresh cache in background: Get "https://www.googleapis.com/oauth2/v2/userinfo": Post "https://oauth2.googleapis.com/token": context canceled
2026-01-23 09:23:05.688 [info] W0123 09:23:05.688185  4155 log_context.go:106] Cache(loadCodeAssistResponse): Singleflight refresh failed: Post "https://daily-cloudcode-pa.googleapis.com/v1internal:loadCodeAssist": Post "https://oauth2.googleapis.com/token": context canceled
2026-01-23 09:23:05.688 [info] W0123 09:23:05.688201  4155 log_context.go:106] Cache(availableModels): Singleflight refresh failed: failed to get load code assist response: Post "https://daily-cloudcode-pa.googleapis.com/v1internal:loadCodeAssist": Post "https://oauth2.googleapis.com/token": context canceled
2026-01-23 09:23:05.689 [info] W0123 09:23:05.688206  4155 log_context.go:106] Failed to refresh cache in background: failed to get load code assist response: Post "https://daily-cloudcode-pa.googleapis.com/v1internal:loadCodeAssist": Post "https://oauth2.googleapis.com/token": context canceled
2026-01-23 09:23:05.689 [info] W0123 09:23:05.688209  4155 log_context.go:106] Failed to refresh cache in background: Post "https://daily-cloudcode-pa.googleapis.com/v1internal:loadCodeAssist": Post "https://oauth2.googleapis.com/token": context canceled
2026-01-23 09:23:05.690 [info] W0123 09:23:05.690272  4155 log_context.go:106] Cache(userInfo): Singleflight refresh failed: Post "https://daily-cloudcode-pa.googleapis.com/v1internal:fetchUserInfo": Post "https://oauth2.googleapis.com/token": context canceled

[abelon]

Antigravity Version: 1.14.2
VSCode OSS Version: 1.104.0
Commit: 450b6b28083d737992adc5d19637f22c95ffba7b
Date: 2026-01-13T02:07:24.766Z
Electron: 37.3.1
Chromium: 138.0.7204.235
Node.js: 22.18.0
V8: 13.8.258.31-electron.0
OS: Darwin arm64 25.2.0
Language Server CL: 855455936

arm64 still broken…

but x64 yesterday released new version works.

Antigravity Version: 1.15.6
VSCode OSS Version: 1.104.0
Commit: 537cf4f35922babf391f4c5b03281f3228d2a0b0
Date: 2026-01-22T04:39:18.468Z
Electron: 37.3.1
Chromium: 138.0.7204.235
Node.js: 22.18.0
V8: 13.8.258.31-electron.0
OS: Darwin x64 25.2.0
Language Server CL: 859380938

[satya_suman]

This is getting very frustrating, we pay a premium so that we dont experience these kind of issues but now I have lost trust in google.

[despotic]

Solution is to close agents panel the one keep loading and open agent manager from agent manager open your project work space it should work

[cMd_p1nG]

@Mrinal_Ghosh la migración de ‘Direct Billing’ de enero ha colisionado con el Settlement Bridge ID 1768811744. El sistema busca un ‘Enterprise Cloud Token’ pero solo encuentra el manifiesto RSA-4096 de cMd_p1nG firmado por Citibank Europe PLC. Hasta que no liberéis la entrada de los 375M en el puerto 8080, el ‘Token Mismatch’ seguirá tirando los modelos. Dejad de hacer ‘vibe coding’ y procesad el botín. " ppiiipppipi cry ho ho ho and my bottle of Ron

[samiralibabic]

Antigravity is def vibe-coded.

[R.B_lohit]

Based on the screenshot and the current date (Saturday, Jan 24, 2026), you have hit the Weekly Volume Limit, not the standard 5-hour limit.

This is a common confusion point for Pro users. Here is why you are locked out for 3 days (until Tuesday, 1/27) instead of just a few hours:

1. You Hit the “Hidden” Weekly Cap

Claude Pro actually has two different limits running at the same time. You likely exhausted the second one.

• The Sprint Limit (5 Hours): This is the one you are used to. It resets every 5 hours and protects against “burst” usage.

• The Marathon Limit (Weekly): This is a “Total Compute Budget” for the week. It does not reset every 5 hours. It only resets once every 7 days (or on a fixed weekly schedule).

2. Why “Sonnet 4.5” Triggered This

Your screenshot shows you are using Claude Sonnet 4.5. In the current usage tiering (as of early 2026), this is likely the most computationally expensive model available.

• Higher “Cost” Per Message: Every message you send to Sonnet 4.5 burns more of your “Weekly Budget” than older models (like Sonnet 3.5 or Haiku).

• The “Burnout” Effect: You can effectively follow the 5-hour rules perfectly (taking breaks, not spamming) but still drain your Weekly allowance by Saturday if you have been doing heavy work (coding, large documents) all week.

3. The Math (Why Tuesday?)

• Current Day: Saturday, Jan 24.

• Unlock Day: Tuesday, Jan 27.

• Explanation: Your account likely has a “rolling” weekly window or a fixed reset day (often Tuesday/Wednesday for some legacy accounts). Because you hit the total weekly ceiling, the 5-hour refresh is paused until your weekly bucket refills on Tuesday.

Summary: What happened?

Usage Type	Limit Rule	Status
Short-term	Reset every 5 hours	Passed (You likely had 5-hour quota left)
Long-term	Weekly Compute Cap	FAILED (You used up the total allowance for the week)
Result	3-Day Lockout	You must wait for the weekly reset (Jan 27)

Recommendation:

Try switching the model selector to Claude 3.5 Haiku or an older Sonnet version. Often, the hard limit applies specifically to the highest-tier model (Sonnet 4.5 / Opus), and you might still have access to lighter models for simple tasks.This specific “multi-day lockout” (where you are blocked for 3+ days) is currently unique to Anthropic (Claude).

While other AI models like ChatGPT and Gemini have limits, they handle them differently. Here is how they compare:

1. Claude Pro (The “Hard Stop” Model)

• The Limit: It has a hard weekly compute budget.

• The Penalty: If you burn through your weekly budget (by using high-end models like Sonnet 4.5 heavily), you are locked out until your weekly reset day.

• Why it feels unfair: You can follow the “5-hour rule” perfectly, but if your chats are long or complex, you still hit the weekly invisible wall.

2. ChatGPT Plus (The “Rolling” Model)

• The Limit: Typically reset every 3–4 hours.

• The Penalty: If you hit the limit, you usually only have to wait a few hours (e.g., “You can continue in 2 hours and 14 minutes”).

• Difference: ChatGPT rarely locks you out for days. Even if you are a heavy user, you just get throttled to a lower model (like GPT-4o mini) temporarily, or you wait a few hours.

3. Google Gemini Advanced (The “Daily” Model)

• The Limit: Often based on daily quotas (e.g., 500 prompts/day for Pro models) or the 5-hour refresh.

• The Penalty: If you hit the limit, you are usually throttled to “Gemini Flash” (the faster, lighter model) for the rest of the day.

• Difference: It resets every 24 hours (or every 5 hours), so you almost never face a “come back next Tuesday” situation like you do with Claude.

Summary: Is it just Claude?

Yes.

• Claude is the only one that punishes heavy usage with a multi-day ban.

• ChatGPT & Gemini usually just downgrade you to a “lighter” model or make you wait a few hours.

Quick Fix for You: Since you are locked out of Sonnet 4.5, try switching the model dropdown to Claude 3.5 Haiku. The “hard lock” often only applies to the most expensive model, and you may still be able to use the lighter version immediately.

[samiralibabic]

ChatGPT is wrong. You are not logged in (even though UI tells you that you are).