Gemini 4 Argon is now available to the general public and developers.

Google is opening Gemini 4 Argon to Pro and Ultra subscribers, developers, and enterprises. Its flagship model is leaving the restricted access phase.

After two weeks of limited access for select partners and cybersecurity specialists, Gemini 4 Argon is now available worldwide. Google is opening its most advanced model to developers, enterprises, and paying subscribers of its services. This rollout comes alongside new demonstrations, particularly in software development, 3D creation, and scientific research. Expanded Access across the Google Ecosystem

Google AI Pro and Ultra subscribers can use Argon in the Gemini app and Google Search's AI Mode. Ultra users can also access it from Google Antigravity, the dedicated development environment for AI agents, available on desktop and command line.

For developers and enterprises, the model is offered in Google AI Studio, via the paid Gemini API, and within Google Cloud's enterprise services.

Launch pricing remains set at $2 per million input tokens and $10 per output token, with a 95% discount on cached inputs. This offer is temporary: starting November 30, announced prices will increase to $4 and $20, respectively. Software Projects Conducted Over Several Weeks

Google is sharing several new internal use cases. On Chromium, Argon helped refactor part of the Chrome browser originating from a legacy codebase of roughly six million lines of code. According to the company, this work, which involved several hundred files, was accomplished in a few weeks rather than several months.

In the open-source Agent Development Kit framework, the model identified optimizations that reduced setup time for long in-memory sessions by 40% and accelerated cold imports by 21%.

Google teams also report an increase of over 35% in code submissions completed using agents in Antigravity, along with a nearly one-third drop in change rollbacks. From Blender to Scientific Research

Visual demonstrations also play a major role in this new presentation. Argon notably uses Blender and Three.js to build interactive 3D environments, create animations, and turn photographs of a desk into a playable miniature racetrack.

In research, Google highlights more specialized applications. The model assisted in optimizing quantum chemistry software and designing new enzymes, some of which subsequently demonstrated catalytic activity during laboratory trials.

Another experiment pairs Argon with Apptronik's humanoid robot, Apollo. From a video showing a person playing cornhole, the model constructed a simulation allowing the robot to learn the required movements. During a real-world test, Apollo successfully landed seven consecutive throws. A Commercial Rollout Under Watch

Broader access does not, however, mean the end of safety restrictions. Google is maintaining safeguards against malicious uses in cybersecurity and biology, as well as monitoring mechanisms for agent-initiated actions.

Concurrently, the Fairwind Program is expanding to several hundred additional partners. Selected cyberdefense professionals can request specific access to the model's capabilities without certain cyber restrictions.

Google also plans to offer 10,000 scientists six months of access to Antigravity's highest usage tier, subject to selection and availability.

With this global launch, Argon shifts from an experimental model reserved for a few organizations to a commercial product. Its extended reasoning capabilities had already been detailed during its initial presentation. The challenge now is to see how they perform in the daily projects of developers, enterprises, and users, beyond Google's curated demonstrations.