Google Introduces Gemini 4 Argon for Coding and Cybersecurity

Google introduces Gemini 4 Argon, a reasoning model for complex coding, professional work, and cybersecurity tasks, as detailed in a company blog post. The model is initially rolling out to trusted cyber defenders through the Fairwind Program, with access for developers, enterprises, paid API customers, and Google AI Ultra subscribers planned later.

Argon supports up to one million output tokens, compared with 64,000 for the previous model. Google reports a score of 77.9% on DeepSWE v1.1 for software engineering tasks and 51.3% on AutomationBench for business workflows.

Google says internal teams use Argon for coding, research, quantum algorithm optimization, and code migrations. The company also reports that Argon agents identified memory optimizations that freed more than 300 TiB across its data centers.

For cybersecurity, Argon can find, validate, and patch software vulnerabilities. Trusted defenders will receive access without cyber guardrails, while broader access will include safeguards against misuse and prompt injection attacks. Introductory pricing will be $2 per million input tokens and $10 per million output tokens, with cached input priced at a 95% discount.