Google Gemini 4 Argon Launches With Cybersecurity Defence CapabilitiesGoogle launched Gemini 4 Argon on 30 September 2026. Google designed this AI model for complex software engineering, enterprise knowledge work and defensive cybersecurity.
Gemini 4 Argon processes up to 1 million output tokens in a single trajectory. This capacity allows the model to execute long-horizon tasks requiring sustained multi-step reasoning. The system analyses professional charts, processes long videos and cross-references multiple documents.
The model scored 77.9 percent on the DeepSWE v1.1 benchmark for software engineering. Google employees currently use the model to debug algorithms and migrate large codebase repositories to the Rust programming language.
Gemini 4 Argon introduces autonomous capabilities for cybersecurity teams. The model finds, validates and patches software vulnerabilities. It achieved a 68 percent score on the CWE-bench v1 cybersecurity benchmark. Security firms currently use the model to identify and remediate exposures in critical infrastructure.
Google restricts initial access to approved cybersecurity partners through the Fairwind Programme. Google removed standard cyber guardrails for these vetted defenders to enable comprehensive vulnerability research. You must secure approval and define authorised systems before deploying the model in your production environments.
Google will expand access to paid API customers and Google AI Ultra subscribers after completing safety evaluations. The company tests the model for chemical, biological, radiological and nuclear risks. Google also implements safeguards against indirect prompt injection attacks.
Google priced the API at $2 per million input tokens and $10 per million output tokens. You receive a 95 percent discount on cached input tokens.
