Google DeepMind Unveils Gemini 4 Argon with 1M Output Tokens for Coding, Knowledge Work and Cyber Defense
Gemini 4 Argon tops GPT-6 Astra and Claude Opus on benchmarks—but Google is keeping it locked down.

Why it matters
Google DeepMind released a frontier model claiming benchmark wins over OpenAI and Anthropic's latest, with 1M output tokens enabling long-form code and analysis work. Access remains gated, leaving practitioners and buyers without independent verification or availability timelines.
The key facts
7 to knowModel: Gemini 4 Argon
1M output tokens (context window/generation limit)
Claimed benchmark lead vs. GPT-6 Astra and Claude Opus 5.5 on most benchmarks
Use cases: coding, knowledge work, cyber defense
Access status: gated (not generally available)
No pricing, regional availability, or rollout timeline disclosed
Source: MarkTechPost (secondary reporting of announcement)
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: Gemini 4 Argon tops GPT-6 Astra and Claude Opus 5.5 on most benchmarks, but access remains gated today.