Google Unveils Gemini 4 Argon AI Model for Developers and Consumers

Google has previewed its Gemini 4 Argon AI model, which will be available to developers and consumers after testing with government agencies and cyber defenders. The model can produce up to one million output tokens in a single trajectory and has stronger long-running reasoning capabilities.

Google is initially providing access to selected cyber defenders through its Fairwind Program and participating in the US government's voluntary process for pre-release model access. When wider access begins, Gemini 4 Argon will start with paid API customers and Google AI Ultra subscribers at $2 per million input tokens and $10 per million output tokens.

The company is already using the model internally, with thousands of employees utilizing it across coding, research, and other workflows. Quantum researchers used Argon to optimize spacetime resources required by computational subroutines, beating a published baseline by 40% in one example.