Google Releases Gemini 4 Argon for Cybersecurity Applications

Google has launched its first model from the Gemini 4 generation, named Argon, focusing on advanced coding capabilities. Initially, access is restricted to trusted cybersecurity professionals and internal teams through Google’s Fairwind Program.

The company announced Argon on September 30th, highlighting its ability to autonomously identify, validate, and patch critical software vulnerabilities. This phased rollout allows Google to gather feedback and refine safeguards before wider distribution.

Key Features & Performance

• Extended output: Generates up to 1 million tokens per response (compared to 64,000 previously)
• Coding proficiency: Scored 77.9% on DeepSWE v1.1 benchmark, surpassing Anthropic’s Claude Opus 5.5 and OpenAI’s GPT-6 Astra
• Vulnerability analysis: Tied with GPT-6 Astra at 68% on CWE-bench v1

Real-World Applications Within Google

Argon is already being utilized internally for:

• Migrating codebases (from C/C++ to Rust)
• Analyzing data center performance and identifying memory optimizations (projected savings of 500 TB - 1 PB)
• Optimizing quantum computing algorithms (demonstrating a 40% improvement in one task)

Security & Pricing

Google’s restricted access reflects the dual nature of powerful cybersecurity tools—both defensive and potentially risky. The company plans to eventually expand availability while maintaining safeguards.

Introductory pricing: $2 per million input tokens, $10 per million output tokens (with higher rates after the introductory period)