Google Unveils Gemini 4 Argon with 1 Million Token Output Limit

Google has announced Gemini 4 Argon, a new AI model designed for long-term tasks in software engineering, cybersecurity, and knowledge work. The model is currently available to vetted cyber defenders through Google's Fairwind Program, with paid API tiers for developers and Google AI Ultra subscribers set to follow.

The main innovation of Gemini 4 Argon is its ability to generate up to 1 million tokens in a single run, exceeding the previous limit of 64,000 tokens. This increased capacity allows the model to tackle complex tasks that require sustained reasoning and context, such as auditing an entire codebase or migrating projects.

Google has already deployed Argon internally across its engineering and infrastructure teams, with notable successes including optimizing Rust code by 2.7 times and freeing over 300 terabytes of memory in server fleet telemetry. On the security front, Argon identified a severe vulnerability in hospital software used worldwide, which earlier models missed entirely.

The phased rollout is tied to the compute requirements and safeguards needed for a model with this level of autonomy, with cyber defenders in the Fairwind Program gaining access before the wider developer market.