Google Launches Gemini 4 Argon with 1M Token Output Limit

The company's new frontier AI model is now rolling out to an initial cohort of cybersecurity partners and internal teams through its Fairwind Program. Gemini 4 Argon specializes in software engineering and cyber defense, achieving a state-of-the-art score of 77.9% on the DeepSWE v1.1 benchmark and competing with rivals like OpenAI's Astra and Anthropic's Opus 5.5. The model expands its output capacity to 1M tokens, a significant increase from 64K in previous versions, and enters the market with introductory pricing of $2 per 1M input tokens and $10 per 1M output tokens.

Google is coordinating with the US government through a voluntary pre-release access process before expanding availability to paid API customers and AI Ultra subscribers. For its internal teams and trusted defenders, the company is deploying the model without standard cyber guardrails, relying instead on misalignment mitigations that monitor the model's chain-of-thought. Internal applications of Argon have already optimized a quantum computing subroutine by 40% and freed between 300TiB and 1PiB of memory across Google's data centers.