Google has started rolling out Gemini 4 Argon, its newest frontier model, to AI Pro subscribers inside Antigravity, the company's agentic development platform, after giving the model to AI Ultra users first. The rollout began on October 15, according to 9to5Google, which confirmed the model is live on its own accounts, and Google announced the expansion on X.
The quota structure is the notable part. Antigravity subscribers get a separate "weekly-only Argon quota" that is consumed on top of the existing allowance for Gemini 3.8 Flash, so the two models coexist rather than compete: Flash stays the high-volume default for everyday work, while Argon is rationed for the hardest tasks. Google has not published how large the weekly quota is, and Pro trial subscriptions are excluded. Reporting on the rollout says Argon is available in the Antigravity app, CLI and IDE extensions, with reasoning levels selectable between Low, Medium and High.
Alongside the expansion, Google published a batch of internal case studies. In one, Argon analyzing production profiling data for YouTube's comments frontend found that redundant emoji rendering alone accounted for nearly 45% of server overhead; Google says it proposed 12 targeted optimizations that cut server comment-load latency by 30 milliseconds and client end-to-end latency by 20 milliseconds. In another, agents cross-checked 134 open bug reports across multiple services and closed 97 stale or duplicate ones in under three hours, shrinking the backlog by 72.4%. Google also says Argon found four critical security vulnerabilities in a batch of access-control changes that two unnamed frontier models had audited and cleared. Every one of these figures is company-reported, comes without a published methodology, and has not been independently verified.
The wider context: Argon was announced on September 30 with an unusual security-first rollout that gave it only to vetted defenders in Google's Fairwind program, then reached paying users broadly on October 13 — AI Pro and Ultra subscribers in the Gemini app and Search's AI Mode, plus developers via the paid API at an introductory $2 per million input tokens and $10 per million output, a rate set to double on December 1. Google says more than 100,000 people inside Alphabet now use the model monthly, that agentic code volume grew 35% in four weeks of internal use, and that post-submit rollbacks fell by 30% — again, self-reported numbers.
Strip away the case studies and this week's move is a distribution change, not a new model. But the structure is telling: Google is treating Argon as a scarce premium resource tied to a paid subscription inside an agent-first coding tool, while Flash absorbs the routine workload. That two-tier pattern — flagship on a weekly leash, cheap model on tap — is quickly becoming the industry's template, and it puts pressure on rivals to publish quota sizes and routing rules so teams can actually plan around them. Until Google says how much Argon a Pro subscriber really gets per week, the practical question for developers remains unanswered.
Comments (0)
Log in to join the discussion
Log InNo comments yet