Google launches Gemini 4 Argon AI model for cybersecurity partners

Google announced Gemini 4 Argon, its most powerful AI model, initially releasing it to select cybersecurity partners through its Fairwind Program before a broader rollout. The model is designed for complex, long-horizon tasks in coding, enterprise work, and cybersecurity defense, and can autonomously find, validate, and patch software vulnerabilities. Google raised the output token limit to 1 million tokens and priced it at $2 per million input tokens and $10 per million output tokens. It said the model is already used internally and in tests, but did not say when it would be publicly available.