Google begins restricted rollout of Gemini 4 Argon
Google is introducing the model to trusted cyber defenders while gathering feedback and strengthening safeguards before wider availability.
AI illustrationGemini 4 Argon is beginning a restricted rollout to trusted cyber defenders before a wider release, Google announced in a September 30 post by Koray Kavukcuoglu. The company said the model is being introduced through its Fairwind Program, with broader availability still to come.
Google said it is participating in the U.S. government’s voluntary process for pre-release model access. Feedback from early testers will inform further guardrail work before the model reaches developers, enterprises, and consumers. The company identified paid API customers and Google AI Ultra subscribers as the starting point for that broader release.
The rollout includes different safeguards for different users. Google said trusted defenders and its internal teams will receive Argon without cyber guardrails. Before wider availability, the company is strengthening defenses against misuse and prompt injection, monitoring model reasoning and actions for misalignment, and hardening testing environments.
Google described a model intended for lengthy coding, research, legal, finance, and cybersecurity tasks. It said the output limit is increasing to 1M tokens from 64K tokens. Introductory pricing will be $2 per million input tokens and $10 per million output tokens, with cached inputs priced at 95% off the input price.
The company also reported a 77.9% result on DeepSWE v1.1, a benchmark for long-horizon software engineering, and a 51.3% score on AutomationBench. These are performance figures presented by Google in the announcement.
Internally, Google said Argon agents are helping migrate C/C++ codebases to Rust. It said large rewrites of critical systems undergo automated and manual audits, emulation testing, and review before production deployment.