Technology

Google Launches Gemini 4 Argon With 1 Million-Token Context for Advanced AI Tasks

Google has announced Gemini 4 Argon, a new frontier AI model from Google DeepMind designed to handle complex, long-running tasks across software engineering, enterprise knowledge work and cybersecurity.

Announced on September 30, 2026, Gemini 4 Argon is initially being rolled out to trusted cyber defenders through Google's Fairwind Program, with broader availability planned for developers, enterprises and consumers.

Gemini 4 Argon gets a 1-million-token context window

One of the biggest upgrades in Gemini 4 Argon is its 1-million-token context window.

Google says this is an increase from the previous 64,000-token limit and is designed to allow the model to work through much longer and more complicated tasks in a single workflow.

A larger context window can be particularly useful when AI systems need to process large codebases, lengthy documents, multiple files or extended sequences of instructions.

Built for advanced coding and enterprise work

Google says Argon has been designed for professional workflows rather than just everyday chatbot interactions.

The company highlights applications including:

Software engineering
Large-scale code migrations
Financial research
Legal research and drafting
Enterprise knowledge work
Chart and document analysis
Long-video understanding
Cybersecurity

Google reports that Argon achieved a 77.9% score on DeepSWE v1.1, a benchmark focused on real-world, long-horizon software engineering tasks. Google also says the model ranked first on its cited AutomationBench evaluation with a score of 51.3%.

These are Google's reported benchmark results and should be considered in the context of the company's testing methodology.

Google is already using Argon internally

Google says Gemini 4 Argon is already being used across internal workflows by thousands of employees.

The company highlighted several examples, including work by its quantum computing researchers to optimise algorithms and efforts to identify memory-saving opportunities across Google's data centres.

Google says Argon agents identified and applied memory optimisations that could free more than 300 TiB of memory once deployed, with potential total savings estimated at between 500 TiB and 1 PiB.

The model is also being used for large-scale software migration projects, including efforts to move C/C++ codebases to Rust.

Gemini 4 Argon focuses heavily on cybersecurity

Cybersecurity is one of the major areas Google is highlighting with Argon.

The company says the model can autonomously identify, validate and patch critical software vulnerabilities.

Google reports that Argon tied for first place on CWE-bench v1, with a score of 68% for vulnerability remediation.

Google also says the model has demonstrated stronger vulnerability discovery capabilities compared with its earlier Gemini 3.8 Flash Cyber model.

Through the Fairwind Program, trusted cyber defenders are getting access to Argon to evaluate its capabilities in real-world defensive cybersecurity situations.

Google is taking a phased approach to the launch

Despite the model's capabilities, Google is not releasing Gemini 4 Argon to everyone immediately.

The company says frontier AI systems at this capability level require additional safety testing and monitoring before broad deployment.

Google is testing protections against misuse, prompt-injection attacks and potential model misalignment. It also says it is hardening the environments in which advanced AI agents operate.

The company plans to expand access gradually after collecting feedback from early testers.

Gemini 4 Argon pricing

Google says Argon will launch at an introductory API price of $2 per million input tokens and $10 per million output tokens.

Cached input tokens will receive a 95% discount during the introductory pricing period.

After the introductory period, Google says the price will increase to $4 per million input tokens and $20 per million output tokens.

When will Gemini 4 Argon be available?

Gemini 4 Argon is currently rolling out to a limited group of trusted cyber defenders.

Google says broader access will begin with paid API customers and Google AI Ultra subscribers, followed by wider availability to developers, enterprises and consumers.

A new direction for Google's Gemini models

With Gemini 4 Argon, Google is positioning its latest model around long-horizon professional tasks rather than simply generating answers to individual prompts.

Its combination of a 1-million-token context window, advanced coding capabilities, enterprise knowledge-work performance and cybersecurity capabilities could make it particularly relevant for organisations using AI agents to complete multi-step workflows.

However, many of the performance figures currently available come directly from Google's own evaluations. Independent testing and broader real-world use will be important in determining how Argon compares with competing frontier AI models.

For now, Gemini 4 Argon represents Google's latest push toward AI systems capable of handling increasingly complex professional and technical work.