Google rolls out Gemini 4 Argon AI model with paid access

Google has commenced the wider rollout of its newest AI model, Gemini 4 Argon, within the Gemini app and the Antigravity platform. Initially announced with a cautious approach due to potential misuse concerns, the AI is now accessible exclusively to paying customers.

Access is not uniform, however, as users on lower-tier paid plans will not receive full functionality. The capabilities of Gemini 4 Argon remain consistent with previous announcements, with Google highlighting its advanced reasoning and agentic coding abilities.

The model is engineered to refuse malicious instructions, supported by an automated misalignment monitor that can halt processing if unintended workflows are detected. Google attributes the necessity for these safeguards to the model's high degree of capability.

Internally, Argon has already demonstrated its utility. Thousands of Google employees have utilized it, contributing to efficiencies such as in YouTube's comment system. The AI identified redundant emoji rendering, the elimination of which reduced server load by 45 percent. Furthermore, Argon assisted in identifying critical bugs within Project Wing's drone physics simulator and has shown measurable improvements in robotics programs.