CODEX RESET RADAR Codex Reset Radar
中文
GUIDE · UPDATED 2026-11-14

Antigravity Usage Limits: Free, Pro and Ultra Quotas Compared

📅 更新 / Updated: 2026-11-14📌 codex-resets.com ↗ 数据档案

Antigravity's limits are a matrix, not a single number: plan tier × model × 5-hour cycle × weekly cap. The official documentation defines the shape; the community documents how it behaves under load. This guide maps the whole matrix so you can tell which constraint is binding before you restructure your work (verified 2026-11-14).

The matrix: plan × model × window§

Every Antigravity session draws from allowances shaped by three variables. Plan tier: a plain Google account gets baseline quotas, Google AI Pro carries larger 5-hour allowances and higher weekly caps, and Google AI Ultra carries the most generous tier. Model: Gemini 3 / 3.1 Pro, Gemini Flash and the Claude models each have their own allowance bucket. Window: each bucket refreshes on the 5-hour cycle until its weekly cap engages.

The official documentation does not publish exact per-model numbers (they move with promotions and load), so the practical reference is the quota display inside the IDE: it shows the remaining allowance per model tier and the countdown to the next refresh. Treat any specific number quoted in forums — including here — as a snapshot, not a promise.

How to tell which constraint is binding§

Three symptoms, three diagnoses. A countdown measured in hours with the label pointing at the model you were using: the 5-hour cycle on that model's allowance. A countdown measured in days: the weekly cap has engaged for that tier. One model blocked while another works: per-model allowances doing exactly what the design says — switch models within the rules and keep going.

The confusing cases reported by the community (timers growing without usage, partial recoveries, 5h→5-day jumps on light use) usually trace back to the weekly cap or to Google-side adjustments rather than your own consumption. The quota display's label and remaining-percentage readout are the fastest way to distinguish "I burned it" from "the system changed".

Making the quotas last§

Keep routine work on Flash-tier models so Pro-model allowances survive for the tasks that actually need them. Batch exploratory runs into the first half of a 5-hour window, leaving a slice for verification. When the weekly cap binds repeatedly, the supported levers are plan tier and workload pacing — not tricks.

Community threads describe third-party API-key side-loading and token-regeneration workarounds; none are documented by Google, some violate its terms of service, and this site does not cover them. For persistent lockouts that contradict the official design, the escalation path that actually works is a support report with timer screenshots.

FAQFAQ

How much quota does a free Antigravity account get?

The official docs describe baseline quotas for free accounts without publishing exact numbers; they are materially smaller than Google AI Pro/Ultra tiers and refresh on the same 5-hour cycle with a lower weekly cap.

Do Claude models inside Antigravity have separate limits?

Yes — community reports and the quota display describe per-model allowances, so Claude-model quota can exhaust while Gemini quota remains. Each bucket refreshes independently.

How do I check my Antigravity quota?

The quota UI inside the IDE shows remaining allowance per model tier plus the countdown to refresh. There is no separate official web dashboard for quota history.

Why do Pro and Ultra users report longer lockouts than free users?

Higher tiers get bigger quotas but also run heavier, longer agent sessions — and community reports cluster multi-day countdowns on Pro/Ultra after sustained use, consistent with the weekly cap engaging. The weekly cap is tier-dependent, so its cooldown dominates at high usage.

🎯 Any confirmed public reset today?
Open the signal radar for sourced announcements; personal allowance and exact reset time remain in Usage.
Open the signal radar →