Antigravity users get Gemini 4 Argon on a weekly quota of its own
Google is rolling out Gemini 4 Argon in its Antigravity coding agent to AI Pro and AI Ultra subscribers on a separate weekly-only quota. Teams now have to decide which tasks justify spending it, and Google's own examples point to sorting them by what a miss would cost.
The Product Desk

A bar showing 97 of 134 open bug reports closed as stale or duplicate by Argon in Antigravity in under 3 hours, according to Google. The remaining 37 were sorted by owner.
- Open bug reports closed, per Google, claim 7 97 of 134 bug reports 72% computed
Open bug reports closed, per Google The rest of the 134
| Share | Part | Whole | Percent | Claim |
|---|---|---|---|---|
| Open bug reports closed, per Google | 97 bug reports | 134 bug reports | 72% (computed) | 7 |
What happened
- Argon is Google's latest flagship model, announced at the end of September and widely rolled out to users earlier this week.
- In a security audit of access-control changes, Argon found four critical bugs in code that two earlier frontier models had each passed as clean, Google says.
- Google says Argon closed 97 of 134 open bug reports in under three hours and sorted the remaining 37 by owner.
- Google says Argon batch-processes tickets from an on-call rotation and has freed up 2-3 hours per shift.
Why it matters
- Who has to decide decision Each team now needs its own rule for which tasks go to Argon, because the picker offers two models with separate budgets.
- What it limits constraint A team that spends its weekly Argon quota early is left with the Flash quota for the rest of the week.
- Who's exposed exposure Teams that route work by Google's examples are relying on internal projects, and the one head-to-head described pits Argon against two unnamed earlier models.
Antigravity's model picker now offers two models and two budgets [4]. It keeps the existing Gemini 3.8 Flash quota and adds Argon beside it, according to Google's post on X [4]. The 9to5Google report does not say how large the weekly allotment is [6].
Google describes Argon as "designed to tackle your hardest problems" [5]. We'd expect many developers to pick the flagship first and keep picking it until the quota is gone, because each person decides alone what counts as hardest.
Google supplied its examples from internal use [12]. They split by whether anyone would notice a miss. In the audit, Argon traced how permissions flowed at runtime, and the two earlier frontier models Google compared it with are not named [8]. The localization job put hundreds of pages into 17 languages in hours, with people doing only the review [10]. The backlog closures came with documented commit proof [7]. That proof gives a reviewer something to check.
The triage example drafts replies for human approval and recurs with the on-call rotation [9]. We'd expect a job like that to draw on a weekly pool far more often than the Fortran-to-Rust rewrite, which Google says gave a 1.8x performance boost [11].
We'd sort each task on two questions: what a miss costs, and whether a reviewer would see it. A missed access-control bug is costly and invisible, because a reviewer cannot check a finding that was never made. A weak translation draft is cheap and visible, because the reviewer reads it.
- Costly and invisible (security audits, access-control review): Argon. - Costly and visible (a migration with a test suite): Flash first, Argon after a failed pass. - Cheap and visible (translation drafts, ticket replies a person approves): Flash. - Cheap and invisible (style cleanups): Flash, with a spot check.
We'd spend Argon only in the first box and start everything else on Flash. The tradeoff is reviewer time, since every Flash miss in a visible box costs a person a retry. If the first box is empty for a week, the Argon quota can go unused.
What to watch
- Whether Google publishes the size of the weekly Argon allotment and whether AI Pro and AI Ultra get different amounts.
- Whether Antigravity shows remaining Argon quota on screen, so a team can see its pace before the week ends.
- Any independent comparison of Argon and Gemini 3.8 Flash on the same Antigravity tasks.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence30
- Adoption
- Not scored
- Hype gap+35
- Incentives70
- Confidence55
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Google is rolling out support for Gemini 4 Argon to Antigravity, its AI coding tool.
- [2]
Gemini 4 Argon is Google's latest flagship AI model, announced at the end of September and widely rolled out to users earlier this week.
- [3]
The Argon rollout in Antigravity is going to AI Pro and AI Ultra subscribers, and 9to5Google saw it live.
- [4]
In a post on X, Google says there is "a separate weekly-only Argon quota" on top of the existing quota for Gemini 3.8 Flash, so users can use the model that best suits their needs.
- [5]
Google says Argon is "designed to tackle your hardest problems."
- [6]
The 9to5Google report does not state how large the weekly Argon quota is.
- [7]
Google says Argon in Antigravity took a queue of 134 open bug reports, closed 97 stale or duplicate bugs with documented commit proof in under 3 hours, and sorted the remaining 37 by owner.
- [8]
Google says Argon in Antigravity audited a batch of access-control changes, traced how permissions flowed at runtime, and found four critical security bugs; two earlier frontier models audited the identical code and both concluded there were zero security holes.
- [9]
Google says Argon in Antigravity batch-processes dozens of tickets from an on-call rotation, proposing fixes and drafting replies for human approval, and has freed up 2-3 hours per shift.
- [10]
Google says its Antigravity DevRel team used Argon to localize hundreds of pages of developer documentation into 17 languages in a matter of hours, shifting human effort exclusively to review.
- [11]
Google says a team used Argon with a custom skill in Antigravity to rewrite a legacy Fortran codebase into Rust, achieving a 1.8x performance boost and memory safety.
- [12]
Google provided its Argon-in-Antigravity examples as internal uses of the model.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
- 9to5google.comGemini 4 Argon is now rolling out to Google’s Antigravity for AI Pro subscribers
1 article · October 15, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- AI Coding AgentsFollow
- AI usage quotasFollow