Mountain View: Google has unveiled Gemini 4 Argon, a new frontier artificial intelligence model designed to handle complex software engineering, enterprise knowledge work and cybersecurity tasks.

The new model is initially being made available to a limited group of trusted cybersecurity defenders as Google carries out additional safety testing before expanding access.

Built for Complex AI Workflows

Gemini 4 Argon is designed to work on lengthy, multi-step tasks that require sustained reasoning rather than short question-and-answer interactions.

Google says the model can support advanced software development, legal and financial research, technical analysis and other professional workflows.

One of its notable features is a 1-million-token output limit, substantially expanding the amount of information the model can process and generate during complex tasks.

Focus on Software Engineering

Coding is one of the major areas targeted by the new model.

Google says its internal teams are already using Argon for debugging, large-scale code migrations and optimisation projects. The company has also reported using the model to assist with migration of large C and C++ codebases toward Rust.

The model achieved a 77.9 percent score on Google’s cited DeepSWE v1.1 benchmark, which evaluates long-horizon software-engineering capabilities.

Stronger Cybersecurity Capabilities

Cybersecurity is another major focus of Gemini 4 Argon.

Google says the model can autonomously identify, validate and patch software vulnerabilities. The company has therefore chosen cybersecurity defenders as some of its first external testers.

Through its Fairwind programme, selected trusted defenders are testing the model while Google continues to evaluate potential risks associated with its advanced capabilities.

Gradual Public Rollout

Google is taking a phased approach to the release.

The company says it is working with the U.S. government’s voluntary pre-release access process and gathering feedback from early testers before expanding availability.

Following the initial testing phase, Google plans to make Argon available to paid API customers and Google AI Ultra subscribers, followed by broader access.

Pricing Announced

Google has announced an introductory price of $2 per million input tokens and $10 per million output tokens.

Cached input tokens will receive a 95 percent discount during the introductory pricing period. Google says standard pricing after that period will rise to $4 per million input tokens and $20 per million output tokens.

AI Development Enters Another Stage

The launch reflects the industry’s continuing shift from conversational AI toward systems capable of completing longer, more complicated tasks with less human intervention.

With its focus on software development, enterprise work and cybersecurity, Gemini 4 Argon is being positioned as a tool for professional and technical environments rather than simply as a consumer chatbot.

Google’s decision to restrict initial access also highlights the growing importance of safety evaluations as increasingly capable AI systems are developed and deployed.