Google rolls out Gemini 4 Argon with free research access for 10,000 scientists

Applications are open for six months at Google Antigravity’s highest usage tier, alongside a global rollout of the model for developers, businesses and paying subscribers

Google’s Gemini 4 Argon rollout includes an application-based offer of six months of free Antigravity access for 10,000 scientists.

Google has begun rolling out Gemini 4 Argon and is offering 10,000 scientists six months of free access to Google Antigravity’s highest usage tier. The research offer accompanies a model designed to handle extended, multi-step work across software engineering, scientific research and professional analysis.

Researchers must apply, with access granted on a rolling basis subject to eligibility, capacity and selection. The offer gives selected scientists access to the development environment Google’s own teams have been using to optimize research software and investigate problems in fields including quantum computing and enzyme design.

Announcing the rollout on LinkedIn, Google CEO Sundar Pichai described Argon as “our latest frontier model built to tackle complex, long-horizon workflows across software engineering and enterprise knowledge work.”

Google’s launch examples range from improving existing code to generating candidates for laboratory testing. The reported results come from Google and early partners, with different evaluation methods and levels of validation.

Research software and university collaboration

One example concerns xTB, an open-source quantum chemistry package scientists use to predict molecular structures and screen potential drugs and materials.

Google says its researchers used Argon in Antigravity to make the package up to 2.3 times faster at predicting molecular structures and 1.8 times faster at searching for alternative molecular shapes. The changes have been contributed to the official xTB code repository.

The company also describes work with Professor Matt Landreman of the University of Maryland on the mathematics behind stellarators, machines being investigated for nuclear fusion.

According to Google, the collaboration identified a broader class of counterexamples to Grad’s conjecture, a longstanding assumption about the possibility of exact solutions for stellarator magnetic fields. The company says the solutions provide benchmarks for checking simulation software used in future designs.

In enzyme engineering, Google reports that Argon assembled data, tuned models and applied filters within hours to produce five new enzyme designs. Subsequent laboratory tests confirmed catalytic activity, according to the company. That example includes physical testing beyond the model’s predictions, although it does not establish performance across enzyme engineering generally.

From document analysis to interactive learning tools

Argon’s launch also includes demonstrations of turning research materials into interactive resources.

Google shows the model combining three open-source repositories into a medical learning tool that lets users explore anatomical layers, clinical information and biomechanics. Another demonstration reconstructs Chaco Canyon in 3D from archaeological records, laser scans, architectural drawings and archives.

These are demonstrations of resource creation, rather than studies measuring whether the resulting tools improve learning.

For professional research, Isomorphic Labs is using Argon to extract information from scientific documents and patents. Google says the work reduces a process that typically takes weeks to hours.

Alex Cabrera, Founding Engineer at early partner Axiom Bio, describes similar applications in experimental analysis: “We've seen it reason through complex experimental data, from cellular imaging to mass spectrometry, and make subtle findings that traditionally require expert scientists to identify.”

Other early evaluations remain at a preliminary stage. Ambience Healthcare reports testing Gemini 4 on more than 200 synthetic longitudinal patient charts and says it is now validating the model on real patient records.

Pricing, safeguards and access

The Gemini API launches with introductory pricing of $2 per million input tokens and $10 per million output tokens. Cached input tokens receive a 95% discount. Google also describes a unified input-output limit of one million tokens for extended tasks through the API.

For work involving multiple steps and tools, Google says it trained Argon to respect system boundaries, report negative findings honestly and seek human confirmation before irreversible actions. Its safety measures include monitoring for unauthorized workarounds and defenses against malicious instructions encountered in external content.

Developers can access Argon through the paid Gemini API, while Google AI Ultra subscribers can use it in Antigravity and AI Studio. Google AI Pro and Ultra subscribers receive access through the Gemini app and AI Mode in Search. Enterprise routes include Gemini for business, the API and Antigravity.

Applications for the scientists’ six-month Antigravity offer are open, with places allocated through the eligibility and selection process.