On October 13, 2026, Google announced the start of general rollout for Gemini 4 Argon, the first fourth-generation Gemini model. Two weeks earlier, on September 30, the company unveiled Argon but granted access only to a narrow circle of cybersecurity specialists. Now the model is open to developers, companies, scientists, and Google AI Pro and Ultra subscribers.
According to Google, Argon is designed for "deep work" — long, multi-step tasks in programming, finance, law, cybersecurity, and science. The model ranks first in the Vals Index, which evaluates AI on tasks from key economic sectors. At the same time, Google increased the response limit to 1 million tokens, and the Gemini app introduced a task mode and a built-in Antigravity environment for macOS.
FinFly examined what is known about the new model, where it leads and where it lags behind competitors, and tested it by building an interactive black hole model in Canvas.
From closed launch to general access
Argon rolled out in two stages. On September 30, Google announced the model and provided it to vetted cybersecurity specialists through the Fairwind program. The company explained this sequence as a security measure: releasing capabilities of this level requires a phased approach. In parallel, Google participated in a voluntary U.S. government procedure that provides access to models before release.
Fairwind appeared in early September alongside Gemini 3.8 Flash Cyber, a version for vulnerability discovery.
Program participants receive Argon without "cyber restrictions" so that defenders can use the model's full potential.
On October 13, the second stage began. The model became available in the API through AI Studio and the Gemini platform, in Antigravity, the Gemini app, and AI Mode in Google Search.
The two-week pause between announcement and release was used to refine protective mechanisms. In its September post, Google promised to gather feedback from early testers and only then open the model to developers, companies, and ordinary users.
How Argon is used inside Google
Google devoted most of the announcement to internal cases. According to the Antigravity team, more than 100,000 Alphabet employees work with Argon monthly in the internal version of the environment. Over four weeks, the share of code written by agents grew by 35%, and the number of rollbacks after submission decreased by 30%.
Science, cybersecurity, and robotics
Beyond software development, Google promotes Argon as a tool for researchers and security specialists. Here the company cites results from several fields.
In fusion energy, Google's team together with Professor Matt Landreman of the University of Maryland found a new class of counterexamples to the 1967 Grad hypothesis about magnetic fields in stellarators. In medicine, Gamow Labs used Argon to analyze the genomes of newborns with rare lung diseases and found a chromosomal rearrangement that standard software had missed for years.
To accelerate research, Google is offering 10,000 scientists free access to the top-tier Antigravity plan for six months. Applications are already being accepted, and participants will be onboarded gradually based on selection criteria and available capacity.
In robotics, Argon scored 87% on the ERQA spatial reasoning benchmark — higher than any frontier model. In a developer blog, Google shows how the model controls Boston Dynamics' quadruped robot Spot, first practicing movements in the MuJoCo simulator. In another test, Argon watched a video of a person playing cornhole, wrote simulation code itself, and Apptronik's humanoid Apollo threw seven bags in a row.
These examples describe the work of teams with early access and expert support. For the average user, they are important primarily as a guide: on which tasks Google expects the model to deliver maximum value.
