Google Rolls Out Gemini 4 Argon, but Only for Paying Users

Google has ended its months-long wait on the frontier. Gemini 4 Argon is now rolling out globally to developers, enterprises and paying subscribers, the company announced on its blog.

What happened

Argon is live in the paid tier of the Gemini API, Google AI Studio, Android Studio and the Gemini platform. Consumer access is gated: the Gemini app and AI Mode in Search require AI Pro ($20/month) or AI Ultra (from $99/month), while the agent-first coding platform Antigravity is Ultra-only, as Ars Technica reported.

Google says Argon ranks first on the Vals Index and ships with a 1M unified input-output token limit plus a new continuation protocol for long-running sessions. Introductory pricing is $2 per million input tokens and $10 per million output tokens, with cached input tokens at 95 percent off. After November 30 those rates double to $4 and $20.

Why it matters

For builders, the number that matters is the price. At $2/$10, Argon lands at roughly a fifth of OpenAI's GPT-6 Astra and Anthropic's Fable 5 models on a per-token basis, according to The Verge. Google is buying adoption for six weeks, then resetting the meter.

The internal usage numbers hint at where agentic workloads are heading. Inside Antigravity, Google reports agentic code submissions up more than 35 percent while code rollbacks fell by nearly a third. An overnight autonomous run against Google's open-source Agent Development Kit cut in-memory session prep time by 40 percent and sped up cold imports by 21 percent.

The catch

Access is the constraint, not capability. Coding-grade Argon is locked to Ultra, there is no timeline for Plus or free users, and the cheapest token window closes on November 30. Google is also keeping Gemini 3.1 Pro around for teams that need more than the Flash models without paying Ultra prices.

Builder impact

Treat Argon's introductory pricing as a migration window, not a baseline. Teams running long multi-step agent loops should benchmark Argon now and decide before December 1, when the same workload costs twice as much. The 95 percent cached-input discount is where long-context agents win, and the 1M unified token limit removes some of the context-stitching plumbing builders currently maintain. Budget for the Ultra tier if Antigravity workflows are on the roadmap.

Read more on how token prices are driving builder costs.