On October 14, 2026, Google officially launched the Gemini 4 Argon model for Google AI subscribers, available on platforms including Google Antigravity, the Gemini app, and Google Search AI mode. The model supports a maximum output limit of one million tokens, priced at $2 per million input tokens and $10 per million output tokens, with prices set to return to $4 and $20 respectively after the initial promotional period. Designed specifically for complex software engineering, enterprise knowledge work, and cybersecurity defense, Gemini 4 Argon is capable of autonomously discovering and fixing software vulnerabilities in over 20 programming languages. It has already been used internally at Google for tasks such as codebase migration and memory optimization. Previously, the model was made available for testing to select cybersecurity experts through the 'Fairwind Program' and outperformed OpenAI's Astra model in multiple benchmark tests.