Gemini 4 Argon Is Google's New Frontier Model for Code and Complex Work

By Previously Staff, Previously

October 14, 2026

Generally available since October 13, Argon tops Google's Vals Index claim, scores 53 on the Artificial Analysis Intelligence Index, and launches at introductory API pricing that doubles on December 1.

Google's newest frontier model is out, and it is aimed squarely at people who build things. Gemini 4 Argon, announced on September 30 and made generally available on Tuesday, October 13, is Google's new flagship for developers, enterprises and paying subscribers — first on the Vals Index by Google's count, built for long, multi-step work across coding, finance, legal, cybersecurity and scientific research, and priced to undercut both OpenAI and Anthropic. It arrives with a deadline: its introductory API prices double on December 1.

The launch was fronted by Koray Kavukcuoglu, Google DeepMind's senior vice president and Google's chief AI architect, who framed Argon as the company's latest frontier model for developers, enterprises, scientists and consumers alike. The rollout is broad: developers and enterprises get it through the Gemini API and AI Studio under the model ID gemini-4-argon; Pro and Ultra subscribers get it in the Gemini app and in AI Mode in Search; Ultra subscribers also get it inside Antigravity, Google's agentic coding platform, with Gemini for Business and Antigravity desktop and CLI access for enterprise teams.

The Numbers Google Is Bragging About

Google says Argon ranks first on the Vals Index, a benchmark that scores models on work across finance, coding, legal and tax tasks. On the independent side, Artificial Analysis scores Argon 53 on its Intelligence Index, against 41 for Gemini 3.8 Flash, the company's fastest current model, according to Fello AI's launch tracker. The technical envelope is serious: a unified context window of 1,048,576 tokens shared between input and output, with default synchronous output capped at 65,536 tokens and a continuation protocol for longer generation, per Google's developer documentation as summarized by Blockchain.News. Argon accepts text, images, video, audio and PDFs as input and produces text output; it does not generate images or audio and does not support the Live API.

The most persuasive evidence is internal. Google reported that engineers using Argon inside Antigravity increased agent-based code submissions by more than 35 percent while cutting code rollbacks by nearly a third, and that the Chrome team used it to complete an architectural refactoring across hundreds of files in weeks — work the company says would otherwise have taken months, The Hindu BusinessLine reported. CEO Sundar Pichai highlighted the model's coding and multi-step tool use in his own launch post. Those are vendor numbers, and we label them as such — but they are specific enough to be checkable over time, which is more than most launch claims offer.

Priced to Poach

The pricing is the strategy. Argon launches at $2 per million input tokens and $10 per million output tokens, with cached input discounted 95 percent to $0.10 — rates that hold only through November 30, before doubling to $4 and $20 on December 1, Google confirmed. Even at the doubled rate, Cryptopolitan noted, Argon matches Anthropic's Claude Opus 5.5 at $4/$20 and undercuts OpenAI's GPT-6 Astra at $10/$50 at introductory prices; on a per-task basis against the Artificial Analysis index, Mashable put Argon's cost at roughly 60 percent of Astra's. The introductory window is a land-grab: seven weeks for developers to build on Argon before the meter doubles, and a migration headache waiting for anyone who integrates deeply and then balks at December's bill.

Tasks, Not Just Chat

For consumers, the visible change is a mode switch. The Gemini app now toggles between chat and tasks: pick tasks, and Gemini can keep working while you are away from your computer, sharing progress as it goes, and connect to more than 40 apps — Instacart, OpenTable, Adobe and Shopify among those Google names — according to AppVolt's walkthrough of the app update. Google also says Argon answers more directly, with less hedging and fewer clichés, and that Antigravity now sits inside the Gemini app for Mac. It is the clearest consumer expression yet of the industry's pivot from chatbots that answer to agents that errand.

The Two-Tier Question

Which brings back the split. Bushletter's reporting points out that paying subscribers' AI Mode answers now run on a different — frontier — model from free users', and that Google has outlined no tools for advertisers or publishers to measure how differently their brands and content surface across the tiers. Okay News added a geographic caveat: Google's own subscription page limits advertised availability to selected countries, so "global" rollout does not yet mean every subscriber, everywhere. Add that Argon's initial September access went to trusted cybersecurity defenders first, per Blockchain.News, and a pattern emerges: this is a carefully rationed frontier, metered by plan, country and use case.

Our Take: A Frontier Model With a Meter Running

Argon looks like the real thing technically — the context window, the tool use, the coding receipts from inside Google itself. But the story of this launch is commercial, not technical: Google has decided the frontier is a subscription feature, priced its API like a Black Friday doorbuster with a December cliff, and quietly made Search itself a two-tier product. Developers should take the introductory window seriously and the December doubling just as seriously. And everyone who uses Google Search for free should understand what changed on Tuesday: the best answers in the building now have a cover charge.