For two weeks, the most powerful model Google has ever built was only available to a small circle of cybersecurity defenders.
That changed on Tuesday. Google made Gemini 4 Argon generally available on October 13, 2026, opening it to developers, businesses and paying Gemini subscribers, and putting a launch price on it that undercuts some of its biggest rivals.
If you pay for Google AI Pro or Ultra, build on the Gemini API, or write code in GitHub Copilot, Argon is either in your hands now or arriving over the next few days.
READ ALSO: Google's New Gemini Agent Gets Its Own Work Email and Runs Claude Too
Gemini 4 Argon is now generally available. The API costs $2 per million input tokens and $10 per million output tokens until November 30, 2026, then doubles to $4 and $20.
From a Locked Room to General Release
Google first unveiled Argon on September 30, but held it back. Only partners in its Fairwind Program, a group of vetted security teams, got early access. According to VentureBeat, Google said it was running extra safety testing and taking part in a voluntary US government pre-release review before a wider launch.
The October 13 release, announced in a post by Google DeepMind's Koray Kavukcuoglu, ends that wait. Google's Gemini API changelog now lists the model, with the ID gemini-4-argon, as generally available. Google describes it as its most capable model for complex coding, agentic, math and science tasks.
Who Can Use Gemini 4 Argon Today
Google split the rollout across its products:
- Gemini app and AI Mode in Search: Google AI Pro and Ultra subscribers.
- Google Antigravity (Google's agentic coding tool): Google AI Ultra subscribers, plus enterprise customers.
- Developers: the paid tier of the Gemini API in Google AI Studio, with a context window of about one million tokens.
- Businesses: eligible Gemini Enterprise plans and Google Cloud.
- GitHub Copilot: Copilot Pro+, Max, Business and Enterprise users, in VS Code, JetBrains IDEs, Xcode, the Copilot CLI and more. GitHub says the rollout will be gradual.
Free Gemini users are not on that list. Google has not given a date for them.
Google is also opening applications for 10,000 scientists to get six months of free access to Antigravity's highest usage tier, subject to eligibility and selection.
What Changes in the Gemini App
For everyday users, Google says Argon answers more directly, with less hedging and fewer repeated phrases, and gives progress updates on longer jobs. The Gemini app also gains a tasks toggle, so Gemini can keep working across your Google apps while you are away from your computer. On Mac, the Gemini app now includes Antigravity's coding interface, so you can switch between chatting and building in one window.
The Price War Angle
The introductory price is the headline for developers. At $2 per million input tokens and $10 per million output tokens, with a 95% discount on cached input, Argon launches well below its standard rate.
VentureBeat's comparison puts that in context. At the launch price, Argon costs about one fifth of OpenAI's GPT-6 Astra and half of Anthropic's Claude Opus 5.5. From December 1, when Argon moves to $4 and $20, it lines up with Opus 5.5's pricing.
Teams that test Argon in the next six weeks get flagship performance at mid-tier prices, then have to decide whether it is worth the full rate.
What Google Claims, and What It Doesn't
Google says Argon ranks first on the Vals Index, which tests finance, coding, legal and tax work, and scores 87% on an embodied reasoning benchmark. On the Gray Swan prompt injection test, Google reports attack success fell to 0.7% over 15 adaptive attempts. Inside Google, the company says engineers using Argon in Antigravity submitted more than 35% more agentic code changes, with code rollbacks down nearly a third.
Those are Google's own numbers. Independent testing will take weeks. Not every outside expert is convinced it leads the field: Sam Witteveen, CEO of Red Dragon AI, told VentureBeat it was a very good model but that he did not think it was the best one available.
On safety, Google says the model was built under its Frontier Safety Framework, with a monitor that watches an agent's reasoning and actions to block unintended workarounds. Security teams can apply to Fairwind for a version without cyber guardrails, which Google is now expanding to hundreds of additional partners.
What This Means for You
If you already pay for Google AI Pro or Ultra, you do not need to do anything. Argon should appear in the Gemini app model picker. If you are a developer, the cheap window runs until November 30, so this is the moment to benchmark Argon against whatever model you use today, using your own prompts and your own data.
For businesses, the bigger signal is competition. Google, OpenAI and Anthropic are now cutting launch prices on their top models, and that pressure tends to bring the cost of serious AI work down for everyone.
READ ALSO: The Best ChatGPT Prompts for Work in 2026: Copy, Paste, Get Results
The takeaway: Gemini 4 Argon is now open to paying users and developers, at a launch price that runs out on November 30. Test it now, before the price doubles.
