Google Releases Its Most Advanced AI Model Gemini 4 Argon, Rolling Out in Stages

Google has released Gemini 4 Argon, its most advanced AI model to date, focusing on long-horizon software engineering, enterprise knowledge work, and cybersecurity defense. The model set records in real-world software engineering tests, tied for first in cybersecurity benchmarks, and leads in professional tasks like finance and law.

Gemini 4 Argon supports up to about 1 million tokens per output, far above the previous 64,000 tokens, making it suitable for large codebases, long documents, and multi-stage tasks. It scored 77.9% on DeepSWE v1.1, beating Claude Opus 5.5's 74.2% and GPT-6 Astra's 74.1%.

Google will roll it out in stages, starting with trusted cybersecurity defenders through the Fairwind Program and participating in the US government's voluntary pre-release model access process. Pricing starts at $2 per million input tokens and $10 per million output tokens, later rising to $4 and $20 respectively.