Gemini 4 Argon
Google expanded Gemini 4 Argon on October 13 from a restricted cyber-defender preview to paid API, Antigravity and Gemini subscriber access.
What happened lately
Gemini 4 Argon timeline
Google Shows Argon Steering Robots From Video and Simulation
Gemini 4 Argon Goes Global: API, Antigravity and Gemini App Access Open
Google unveils Gemini 4 Argon with access limited to cyber defenders
A restricted frontier-model rollout
Google announced Gemini 4 Argon on September 30, positioning it for long-running software engineering, professional knowledge work and cybersecurity defense. Initial access went to selected cyber defenders through its Fairwind Program. On October 13, Google announced a wider global rollout through the paid Gemini API, eligible Google Antigravity accounts, and Google AI Ultra and Pro subscriptions in the Gemini app and AI Mode in Search. See the October 13 access report for the current routes and conditions.
Capabilities and safeguards
Google says Argon can produce up to one million output tokens in a single trajectory, compared with a prior 64,000-token limit. It reports strong results on coding and enterprise benchmarks and describes internal uses for code migration and vulnerability remediation. Those are company-reported evaluations and examples, not independent proof that the model will work equally well in other environments. Google says it is testing safeguards for cyber misuse, prompt injection and unintended agent actions before a wider rollout, while giving selected defenders access without its usual cyber guardrails.