Trusence Every claim has a source
Last updated 17 November 2026 Search Türkçe
Models

Google launches Gemini 4 Argon, starting with trusted cyber defenders

Google is limiting early access to a model built for high-risk work before opening it more widely later.

Updated 4 October 2026

Google has launched Gemini 4 Argon, its new frontier model, but is opening access only in stages because it is tuned for high-risk work like cybersecurity, code migration, and other long-horizon tasks. The model is already in use inside Google and by a small set of trusted cyber defenders, while broader availability will come later for developers, enterprises, and consumers. Google says Argon can autonomously find, validate, and patch critical software vulnerabilities, and it is being used in internal projects that include quantum research and memory optimization across data centers. The company also says Argon starts at $2 per million input tokens and $10 per million output tokens, runs with a 1 million token output limit, and has strong benchmark results on software engineering and vulnerability remediation.

Why it matters

Argon’s first users get a model that can already be used on vulnerability work and internal operations, while everyone else has to wait for a wider release. That split suggests Google is treating it as a tool for sensitive, long-running tasks rather than an immediate general rollout, with pricing and a 1 million-token output limit set for when access expands.

Sources · 5