Gemini app rolling out Gemini 4 Argon to AI Pro and AI Ultra

Google is rolling out Gemini 4 Argon to the Gemini app, Google Antigravity, and more, after an announcement at the end of September. Abner Li's 9to5Google report, dated October 13, 2026 at 9:00 am PT, says Google AI Pro and AI Ultra subscribers both get the model in the app. Last month only the Ultra tier, at $99.99 and up, had access. The model picker shows 3.5 Flash-Lite, 3.8 Flash, 3.1 Pro, and 4 Argon, and both tiers can use Gemini 4 Argon in AI Mode.

This piece covers the access gates, the macOS app changes, Chrome, Gemma, and YouTube results from Google's internal Antigravity, and the Gemini API price until November 30, 2026. It is for AI Pro and AI Ultra subscribers, Antigravity, AI Studio, and Android Studio Canary developers, API customers, and scientists applying for access.

Gemini app rolling out Gemini 4 Argon to AI: the announcement

Google says to expect "frontier reasoning and improved performance across complex workflows" in the Gemini app. The company says Argon "answers questions more directly with less hedging, doesn't repeat itself more than necessary, and skips clichés throughout the responses." It "follows style instructions more reliably than any of our previous Gemini models." For longer tasks, Google says, "it keeps you in the loop by sharing its progress more often and more effectively." That is the behavior description attached to moving Argon past last month's Ultra-only access.

Google cites "major improvements" in coding performance and is letting users "switch between chatting and coding" in the Gemini app for macOS. Users get direct access to Antigravity's agentic coding interface in the consumer client, for building apps, web experiences, and more. A new toggle in the main chat window switches from chat to tasks, a mode previously named Spark. AI Pro and AI Ultra still see 3.5 Flash-Lite, 3.8 Flash, and 3.1 Pro in the picker beside 4 Argon.

What actually changed with Gemini app rolling out Gemini 4 Argon to AI

Google shared examples from its internal version of Antigravity across Chrome, Gemma, YouTube, and more. Argon helped the Chrome team separate core webpage handling from a roughly 6 million line legacy folder of browser features inside Chromium's 40 million-line codebase. Piping multiple clean code changes at a time, it helped finish an architectural refactor across hundreds of files, completing months of work in weeks. For Chrome's billions of global users, Google says live webpages will move fluidly between tabs, side panels, and floating windows without reloading.

On Gemma, Google said Argon delivered a 2.85x end-to-end inference speedup on an A100 via vLLM, raising throughput from about 37,900 to over 108,000 tokens per second on Gemma 4 E2B. The run kept 100% token parity and zero quality drop. Google said the work took a few days, versus a process that would usually take months. Upstream pull requests record the changes, from attention tuning to speculative decoding, the same internal Antigravity effort Google grouped with the Chrome and YouTube examples.

Who should care about Gemini app rolling out Gemini 4 Argon to AI

AI Pro subscribers are the people this rollout adds in the Gemini app. Last month only AI Ultra, at $99.99 and up, had Gemini 4 Argon. Both AI Pro and AI Ultra now have it in the Gemini app and in AI Mode. The Antigravity desktop app and the Antigravity CLI still require AI Ultra, and so do AI Studio and Android Studio Canary. On macOS, subscribers can switch between chatting and coding through Antigravity's agentic coding interface in the consumer client.

YouTube is the production case for teams watching CPU and latency. Argon in Antigravity cut the CPU footprint of YouTube's comments frontend in half. Tracing production profile data, Argon found that redundant emoji rendering alone accounted for nearly 45% of server overhead, leading to 12 targeted optimizations. Server comment-load latency fell by 30 ms, and client end-to-end latency fell by 20 ms. Google is also offering 10,000 scientists the highest usage tier in Google Antigravity free of charge for 6 months.

How to try Gemini app rolling out Gemini 4 Argon to AI

In the Gemini app, AI Pro and AI Ultra subscribers choose 4 Argon in the model picker. As of October 13, 2026, that picker also lists 3.5 Flash-Lite, 3.8 Flash, and 3.1 Pro. Both tiers can open Gemini 4 Argon in AI Mode. In the Gemini app for macOS, a subscriber can switch between chatting and coding and use Antigravity's agentic coding interface in the consumer client for apps, web experiences, and more. The toggle in the main chat window moves a session from chat to tasks, the mode previously called Spark.

The Antigravity desktop app and the Antigravity CLI require AI Ultra for Argon. AI Studio and Android Studio Canary require AI Ultra as well. On the Gemini API the introductory price is $2 per million input tokens and $10 per million output tokens until November 30, 2026, then $4 per 1M input tokens and $20 per 1M output tokens. Applications are open for 10,000 scientists to receive the highest usage tier in Google Antigravity free of charge for 6 months.

What to watch after Gemini app rolling out Gemini 4 Argon to AI

The Gemini API price changes on November 30, 2026. Until then Argon costs $2 per million input tokens and $10 per million output tokens. After that date the price is $4 per 1M input tokens and $20 per 1M output tokens. Separately, 10,000 scientists can get the highest usage tier in Google Antigravity free of charge for 6 months, and applications are open now. The Antigravity desktop app, Antigravity CLI, AI Studio, and Android Studio Canary stay on AI Ultra when AI Pro includes Argon in the Gemini app and AI Mode.

Google says the Chrome work will unlock live webpages that move fluidly between tabs, side panels, and floating windows without reloading, eliminating an entire category of browser crashes for billions of global users. The refactor separated core webpage handling from a roughly 6 million line legacy folder in Chromium's 40 million-line codebase and covered hundreds of files, completing months of work in weeks. Gemma 4 E2B's 2.85x speedup on an A100 via vLLM is already tied to upstream pull requests from attention tuning to speculative decoding.