Gemini 4 Argon Lands in the Gemini App for AI Pro Subscribers. Here’s When to Use It

Two weeks ago, Gemini 4 Argon was a model most people could only read about. Google announced it on September 30 and handed it first to cybersecurity defenders in its Fairwind Program, then to AI Ultra subscribers paying $99.99 a month or more. As of October 13, it’s in the Gemini app for Google AI Pro subscribers too, the $19.99-a-month tier that many Android users already pay for, or got as a trial with a new phone.

That’s the moment Argon stops being a benchmark story and becomes a “should I care” story. Short answer: if you already pay for AI Pro, yes, switch to it today. If you’re on the free tier or AI Plus, nothing here changes for you yet, and you shouldn’t upgrade just for Argon until you know what you’d use it for.

What Gemini 4 Argon is

Argon is the first model in the Gemini 4 family. Koray Kavukcuoglu, Google DeepMind’s SVP and chief AI architect, pitched it as a frontier model for software engineering, enterprise knowledge work and cyber defense. Google’s own numbers lean hard on long, multistep work: 77.9% on the DeepSWE v1.1 software-engineering benchmark, 91.7% on the LVBench long-video test, and first place on Zapier’s AutomationBench at 51.3%.

Independent testing backs up the general picture. Engadget, citing Artificial Analysis, reported that Argon matches OpenAI’s GPT-6 Astra on its Intelligence Index at about 60% of the cost per task, and posted a 15% hallucination rate, the lowest among the leading models in that test. One number that matters for real use: Argon can produce up to 1 million output tokens in a single response, up from 64,000 on Google’s earlier models.

Benchmarks are marketing until they show up in your chats, though. The changes Google highlights for the Gemini app are more mundane and arguably more useful.

• More direct answers. Google says Argon hedges less, repeats itself less and leans on fewer clichés.
• Better at following style instructions. Tell it “no bullet points, keep it under 100 words,” and it’s supposed to listen.
• Progress updates on long jobs. On bigger requests it reports where it is more often and more clearly.
• Stronger coding. On Mac, the Gemini app now builds in Google Antigravity’s agentic coding interface, so you can flip from chatting to coding in one window.

Who gets Argon, and where

This is where the rollout gets messy. Argon isn’t available everywhere a paid subscriber might expect, and the coding tools are still reserved for the top tier.

Google’s October 13 post names only Pro and Ultra for the Gemini app, so Plus subscribers are on the outside for now. Usage limits for each tier haven’t been published.

How to switch to Argon on your Android phone

1. Update the Gemini app from the Play Store. The rollout is gradual, so it may take a day or two to reach your account.
2. Open a new chat and tap the model picker.
3. Choose 4 Argon. Pro and Ultra subscribers now see four options: 3.5 Flash-Lite, 3.8 Flash, 3.1 Pro and 4 Argon.

The same release adds a toggle at the top of the main chat window that switches between chat and tasks, the agent feature Google used to call Spark. Tasks can keep running while you’re away from your computer, and Gemini can now connect to more than 40 apps, including Instacart, OpenTable, Adobe and Shopify.

When to use Argon, and when not to

Don’t make Argon your default for everything. A bigger reasoning model is slower, and for “what’s a good substitute for buttermilk” or a quick email rewrite, 3.8 Flash will feel snappier and do the job. Where Argon earns its spot is the stuff that used to fall apart halfway through: comparing three long PDFs, pulling specific moments out of an hour-long video, planning a multistep trip with real constraints, or debugging a script that’s been fighting you all afternoon.

Google’s internal examples point the same direction. The company says Argon helped split a roughly 6-million-line legacy folder out of Chromium’s 40-million-line codebase in weeks instead of months, and helped YouTube’s team halve the CPU footprint of its comments frontend. You won’t do either of those in a chat window on your Pixel. They do tell you what the model is built for: long, tangled jobs where older models lose the thread.

The developer price is the quiet headline

If you build apps, the API pricing matters more than the consumer rollout. Argon costs $2 per million input tokens and $10 per million output tokens through November 30, then rises to $4 and $20. Cached input is 95% cheaper than standard input. Engadget put that next to GPT-6 Astra at $10 input and $50 output, which makes Argon one-fifth the price during the intro window and still less than half after it ends.

Consumer subscriptions don’t include API usage, which is billed separately.

The verdict

Google got the order of this rollout right. Putting Argon in the $19.99 Pro tier instead of keeping it locked behind Ultra means far more paying Android users get the company’s best model without a fivefold price jump. The catch is that the most impressive tricks, agentic coding in Antigravity and AI Studio, still sit behind Ultra. For most Pro subscribers, the real upgrade is simpler and more welcome: a Gemini that talks straighter and finishes long jobs instead of wandering off. Switch the model picker and give it your hardest problem first.

Sources: Google, 9to5Google, Engadget, Google DeepMind. Photo by Anastasiia Nelen on Unsplash.