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)
The story so far
Earlier coverage of this storyline
- Gemini 4 is almost ready, says new Google DeepMind chiefThe Verge AI
- Google plans Gemini 4 release before year-endComputerworld
- Gemini 4 Argon: our next era of frontier intelligenceGoogle DeepMind Blog
- Google rolls out Gemini 4 Argon, its most advanced AI modelCNBC Technology
- Google announces Gemini 4 and says it’s so capable that only ‘trusted cyber defenders’ can have it right nowThe Verge AI
- This story
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.