Google unveils Gemini 4 Argon, its most advanced model yet
What's the story
Google has unveiled Gemini 4 Argon, its most advanced artificial intelligence (AI) model to date. The new system is designed to perform a range of tasks, including coding, research, and writing. However, Google's primary focus with this release is on cybersecurity capabilities. The tech giant says that Argon can "autonomously find, validate, and patch critical software vulnerabilities."
Cybersecurity focus
Argon is being rolled out to select partners
Gemini 4 Argon is being rolled out to select partners through Google's Fairwind Program, a security initiative.
The model was specifically trained for defensive cyber work and has already been used by Google employees for tasks like debugging and codebase migrations.
Beyond cybersecurity, Argon also excels at parsing visuals such as long videos or charts.
Performance comparison
Gemini 4 Argon outperforms GPT-6 Astra, Fable, and Opus
Google has touted the performance of Gemini 4 Argon against models like OpenAI's GPT-6 Astra and Anthropic's Fable and Opus.
The company claims that Argon scored significantly higher than these competitors across a range of AI benchmarks.
This comes as part of a wider trend in which leading AI labs are racing to release more powerful models, despite warnings about potential risks associated with advanced artificial intelligence systems.
Twitter Post
Take a look at the benchmark scores
Introducing Gemini 4 Argon – our new frontier model.
— Google DeepMind (@GoogleDeepMind) September 30, 2026
It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program. pic.twitter.com/X8acOWJOSF
Advanced features
Model has output token limit of 1 million tokens
Gemini 4 Argon boasts an output token limit of one million tokens, a massive increase from the previous limit of 64,000. This allows for longer and more complex use cases.
The model also comes with enhanced coding, reasoning, and multimodal capabilities.
Google plans to make it available "without cyber guardrails" to trusted defenders and internal teams so they can leverage its full cybersecurity defense capabilities.
Domain expertise
Argon excels at specialized tasks
Beyond coding, Gemini 4 Argon has also shown top-notch performance in other domain-specific evaluations.
These include the Vals Finance Agent v2 for multi-step financial research and Harvey's Legal Agent Benchmark for legal research and drafting.
The model is also being used internally at Google, with thousands of employees highlighting its strengths in specialized coding tasks, deeper research, and quality writing.