AI & Tech Unverified First seen 2026-09-30 23:20 UTC

Google releases Gemini 4 Argon AI model, leads cybersecurity benchmarks

Decrypt reports that Google's Gemini 4 Argon tops 12 of 18 benchmarks in Google's own table, writes a million tokens per reply and resists hijacking best, with cyber defenders getting it first.

Why “Unverified”

Reported by Decrypt alone, with no second source or primary confirmation yet

1 independent source groups across 1 posts · evidence score 56/100. How labels work

What the sources claim

Gemini 4 Argon tops 12 of 18 benchmarks in Google's own table.

“Gemini 4 Argon tops 12 of 18 benchmarks in Google's own table”
Decrypt
Gemini 4 Argon writes a million tokens per reply.

“writes a million tokens per reply”
Decrypt
Gemini 4 Argon resists hijacking best.

“resists hijacking best”
Decrypt
Cyber defenders get Gemini 4 first, with the guardrails off.

“Cyber defenders get it first, with the guardrails off”
Decrypt

Sources

Decrypt newswire 2026-09-30 23:20 UTC
Gemini 4 Is Here, and Google’s Flagship Tops All Other AI Models on Cybersecurity

Summary and claims extracted by deepseek/deepseek-flash with prompt analyze-story@3 on 2026-09-30 23:30 UTC. Labels are computed from the sources, not by the model.