Google’s new Gemini 4 Argon targets complex coding, finance and legal work

Google has launched Gemini 4 Argon, its latest frontier artificial intelligence (AI) model, with a focus on complex reasoning, long-running tasks, software engineering, and enterprise work.

In its announcement, Google said that the model is designed to handle workflows that require AI systems to work through multiple steps rather than simply answer individual prompts. Google is initially rolling it out to a group of trusted cybersecurity defenders through its Fairwind Program, before making it available more broadly to developers, enterprises and consumers.

A much longer output window


One of the biggest changes in Argon is the amount of information the model can process and generate in a single trajectory.

Google has increased its output token limit to 1 million tokens, up from 64,000 tokens previously. This gives Argon considerably more room to reason through complex problems, write large amounts of code or work through lengthy research tasks without having to break them into multiple interactions.

Also Read: FTC opens probe into AI giants including Anthropic and OpenAI, source says

Google says this is particularly useful for long-horizon tasks where an AI agent may need to plan, execute, check its work, and continue working over many steps.

Coding and software engineering

Google says the model is already being used internally by thousands of employees for debugging, algorithm design, and large-scale code migrations. It is also working on migrating C and C++ codebases to the Rust programming language, including projects ranging from tens of thousands of lines of code to more than 800,000 lines in the Fuchsia Zircon kernel.

Beyond coding

Google is positioning Argon as an enterprise model rather than simply a programming assistant. The company says that it leads the Vals Index, which measures AI performance across areas including finance, coding, legal, and tax work.

The model also has strong multimodal capabilities. Google says Argon can analyse professional charts, understand long videos, and use information across multiple documents.

Built for cybersecurity

Google said the model can autonomously find, validate, and patch critical software vulnerabilities. The tech giant added that it is also working with cybersecurity company Wiz, which is using Argon through its Scan for Good initiative to identify and remediate exposures in critical infrastructure.

For trusted cybersecurity defenders and its own internal teams, Google says Argon will be released without cyber guardrails so they can use its full capabilities for defensive work.

Also Read: OpenAI launches Dots to take on Meta’s Muse with AI agents that can act on your behalf

What is Google doing about safety?

The increased autonomy of Argon also means Google is putting additional safeguards around the model before a wider rollout.

The company says it is strengthening protections against misuse, including cyber and chemical, biological, radiological, and nuclear risks. It has also worked on making the model more resistant to indirect prompt injection attacks, where malicious instructions embedded in external content can attempt to manipulate an AI agent.

Google is also deploying systems to monitor Argon's reasoning and actions for signs of misalignment — essentially, behaviour that goes beyond what the user intended. The system can stop execution when necessary.

The company says similar monitoring was used during training runs, with potential incidents sent to a dedicated response team. Google says it deliberately avoided feeding those findings back into training to reduce the risk of the model learning to evade the monitoring.

When will Gemini 4 Argon be available?

For now, Argon is being released in phases, starting with trusted cyber defenders and testers. Google says it is participating in the US government's voluntary pre-release model access process while using feedback from early users to improve its safeguards.

The model will eventually be available to developers, enterprises and consumers, starting with paid API customers and Google AI Ultra subscribers.

Google is pricing Argon at an introductory $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% below the standard input rate.