Google has introduced Gemini 4 Argon, calling it its most powerful AI model yet. The model is currently available only to selected users and organisations.
Built for Complex Tasks
Argon is designed for advanced reasoning, software engineering and professional workflows. Google says its teams are already using it for programming and quantum computing.
One Million Token Capacity
Gemini 4 Argon supports an output limit of up to one million tokens. This allows the model to handle lengthy and complex tasks in a single sequence.
Strong Benchmark Performance
Google says Argon delivers competitive results against leading AI models. It recorded 68.9% on the Vals benchmark for knowledge-based tasks.
Focus on Coding
Argon achieved 91.9% on the Vibe Code Bench programming test. Google says the model is built to handle demanding software-development workflows.
Cybersecurity Capabilities
Argon has been trained to identify, validate and fix critical software vulnerabilities. Google says it can support cybersecurity defenders in finding previously missed threats.
Wiz Puts Argon to Work
Google-owned cybersecurity company Wiz is testing Argon through its “Scan for Good” initiative. The model reportedly identified a critical vulnerability in healthcare software.
Coming to Paying Customers
Broader access will begin with eligible API customers and Google AI Ultra subscribers. Google says it plans to expand availability while continuing to improve safety measures.
Gemini 4 Argon Pricing
The introductory API price is $2 per million input tokens and $10 per million output tokens. After the launch period, pricing will rise to $4 and $20 respectively.