Google rolls out Gemini 4 Argon to Antigravity for AI Pro and Ultra subscribers with separate weekly quota

Google is rolling out Gemini 4 Argon to Antigravity, its agentic development platform, for AI Pro and AI Ultra subscribers. 9to5Google reports the model is live on its own accounts. Antigravity users get a separate "weekly-only Argon quota" on top of the existing quota for Gemini 3.8 Flash, according to a Google post on X.

What is confirmed

• Model: Gemini 4 Argon is Google's latest flagship model. It was announced at the end of September 2026 and reached users broadly earlier this week.
• Product: Antigravity is Google's AI coding tool. Google describes it as an agentic development platform that uses AI and agents to help build apps.
• Eligibility: AI Pro and AI Ultra subscribers.
• Quota structure: Argon has its own weekly-only quota. Gemini 3.8 Flash keeps its existing quota, so users can pick the model that fits the task.
• Pricing, context window, benchmark scores and parameter count: not disclosed in this announcement. API pricing for Argon is not yet disclosed here.

Google says Argon is "designed to tackle your hardest problems."

Google's internal examples (company claims)

Google published a set of internal use cases. None have been independently verified, and Google gave no methodology.

• YouTube comments frontend: Google claims the CPU footprint was cut in half. Argon reportedly traced production profile data and found redundant emoji rendering accounted for nearly 45% of server overhead. It led to 12 optimizations, a 30 ms drop in server comment-load latency and a 20 ms drop in client end-to-end latency.
• Bug backlog: Of 134 open bug reports, Google says agents closed 97 stale or duplicate ones with commit proof in under 3 hours, a 72.4% queue reduction. The remaining 37 were sorted by owner.
• Security audit: Google says Argon found four critical security bugs in a batch of access-control changes. It says two earlier frontier models audited the same code and found none. Those models are not named.
• Fortran to Rust migration: Google claims a complete rewrite of a legacy scientific codebase with a 1.8x performance gain.
• Wing drone simulator: Argon proposed 18 modelling improvements across aerodynamics, propulsion and structural dynamics.
• Localization: Hundreds of documentation pages were translated into 17 languages in hours.
• Data center analytics: A 3,200-line fleet analytics pipeline (Python, SQL, dashboards) reached production in days.
• Sales qualification: Google claims a 90% cut in research time, saving upwards of 10 hours per week.
• Finance: Across 80 workflows over seven business days, Google says 15.5 hours of human direction produced about 335 human-equivalent hours of validated output.
• Support triage: Google says the workflow frees 2-3 hours per shift for an on-call team.

What this means

This is a distribution change, not a new model. Argon was already out, and Antigravity is now another place to use it. The separate weekly quota matters more in practice. Google is treating Argon as a scarce, premium resource and keeping Gemini 3.8 Flash as the high-volume default. Developers should expect to ration Argon for hard tasks like audits, migrations and cross-service debugging, and use Flash for routine work.

The case studies are marketing material from Google's own teams. They are useful for seeing the workloads Google wants Argon associated with: long-running agentic tasks, runtime tracing and multi-step cross-referencing. They are not a substitute for independent evaluation. The security audit claim in particular rests on unnamed comparison models, and the reported productivity figures have no published baselines.

The missing weekly quota size is the key gap. Until Google publishes it, teams cannot judge whether Argon in Antigravity can support sustained agentic workloads or only occasional use.