Google unveils Gemini 4 Argon for coding, enterprise and cybersecurity work

Google has announced Gemini 4 Argon, its newest AI model designed to handle complex, long-running tasks.

The model features a one million-token output limit, up from 64,000 tokens in the previous model, which Google said allows it to handle longer, more complex use cases across software engineering, including large-scale code migrations, as well as enterprise work such as financial research, legal research and drafting. It can also analyse charts and long videos and take action based on information across multiple documents.

“Built to sustain deep reasoning across complex, long-horizon workflows, Argon is fundamentally changing the way we work and build at Google,” wrote Google Chief AI Architect and DeepMind SVP Koray Kavukcuoglu in a blog post.

Kavukcuoglu highlighted the model is already being used internally by employees for specialised coding tasks, research and writing. In one example, Argon helped Google’s quantum computing researchers improve the resources used by quantum algorithms, beating a published baseline by 40% “in a matter of minutes”.    

In cybersecurity, Kavukcuoglu said Argon can autonomously find, validate and patch critical software vulnerabilities. Security company Wiz is already using Argon through its Scan for Good initiative, which aims to identify and remediate high-risk exposures in critical public infrastructure. 

More critical vulnerabilities exposed by Google Gemini 4 Argon

“In an early demonstration of its impact, the model uncovered a critical vulnerability exposing sensitive personal information across healthcare software used by hospitals worldwide, identifying a severe risk that previous frontier models had missed,” Kavukcuoglu said. 

According to Kavukcuoglu, Google is taking a phased approach to the model’s release. It’s being initially made available to a group of cyber defenders through Google’s Fairwind Program. 

“Safely releasing frontier capabilities at this level requires a phased approach,” he said. 

Kavukcuoglu added Google is participating in the US government’s voluntary process for pre-release model access and is continuing to test and strengthen safeguards. 

He said there are plans to make Argon available to developers, enterprises and consumers as testing progresses, starting with paid API customers and Google AI Ultra subscribers.

“We’ll continue to gather feedback from early testers as we iterate on guardrails before making Argon available to developers, enterprises, and consumers as soon as possible,” he said. 

Argon will launch at an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% below the standard input token price.

You might also be interested

• UK forecast cloudy as Microsoft, Google and others announce AI data centres
• AMD to invest up to $5 billion in Anthropic
• OpenAI turns 600 employees into millionaires: Inside the mega deal
• Forget AI, here’s the real new trend Wall Street is missing