Google rolls out Gemini 4 Argon for enterprise users

Google has rolled out Gemini 4 Argon to paying developers, enterprise users and subscribers across its AI products. The model is now available through the paid tier of the Gemini API, Google AI Studio, the Gemini app and AI Mode in Search.

Google is positioning Argon for complex, multi-step work in coding, finance, legal, cybersecurity and science. Internal teams, early enterprise users and researchers have already been using the model in production settings and research projects.

In software engineering, Argon has increased agentic code submissions by more than 35% while cutting code rollbacks by nearly a third in recent weeks, according to Google. One example involved the Chrome team, which used the model to separate core webpage handling from a legacy folder of browser features in Chromium.

That task involved a codebase of about 40 million lines, including a legacy folder of roughly 6 million lines. Work that would normally have taken months was completed in weeks, Google said.

Enterprise use

Google named Deloitte, LexisNexis, Ambience Healthcare and Isomorphic Labs among early access users. Customers and start-ups have been using Argon for deep research, document synthesis, longitudinal data analysis and the analysis of large codebases, it said.

The release reflects Google's effort to place a more specialised model across a broad set of commercial channels as large technology groups compete for developer spending and corporate AI budgets. By making Argon available through both APIs and consumer-facing subscription products, Google is widening the ways customers can test and adopt the model.

Cybersecurity push

Google is also expanding its Fairwind Program to hundreds of additional partners after earlier testing. It linked that move to cybersecurity work carried out with Wiz through its Scan for Good initiative.

According to Google, Wiz used Argon to chain complex bugs and patch critical flaws affecting hundreds of thousands of private patient records. Google presented that work as an example of how the model can support security investigations that require multi-step reasoning across large technical systems.

Research work

Outside commercial deployments, the model has also been used in scientific research, according to Google. The company cited work on quantum error correction through AlphaQubit and mathematical research on clean fusion with a Professor at the University of Maryland.

In one of those projects, Argon nearly doubled quantum error-correction throughput, Google said. The company did not provide broader benchmark figures in the announcement, but said the model outperformed other frontier systems across a range of domains.

The launch expands Gemini's role across Google's software and subscription products as the company seeks to convert advances in model development into paid usage. It also shows how major AI providers are increasingly pointing to evidence from coding teams, security researchers and professional services groups to demonstrate value in practical tasks, rather than benchmark scores alone.

Access has been split across product tiers. Developers and enterprise customers can use Argon through the paid Gemini API, while subscribers to Google AI Ultra and Pro can access it through selected consumer products, including the Gemini app and AI Mode in Search.

For business users, the significance lies less in a single product launch than in the breadth of distribution and the choice of target workloads. Google is directing Argon at areas where customers tend to pay for accuracy and time savings, particularly software development, research-heavy document work, regulated professional tasks and security operations.

Google said Argon has already been used to complete months of engineering work in weeks and to help patch flaws tied to hundreds of thousands of private patient records.