Google DeepMind

AI · Cybersecurity · Enterprise Software

Launch

Google DeepMind launches Gemini 4 Argon, 1M-token model, for cyber defenders

Routing its most capable model to vetted security teams before consumers reveals how seriously Google treats frontier AI as a potential cyberweapon.

• Google DeepMind unveiled Gemini 4 Argon on September 30, 2026, rolling it out first to vetted cyber defenders through its Fairwind Program before wider release.
• Argon is Google's new flagship frontier model, built to sustain deep reasoning across long-horizon software engineering, legal and finance work, and cybersecurity defense.
• Google raised the model's maximum output to 1 million tokens, up from 64,000 on prior Gemini models, letting it reason through complex problems in a single continuous pass.
• Argon scored 77.9% on the DeepSWE v1.1 coding benchmark versus 74.2% for Anthropic's Claude Opus 5.5 and 74.1% for OpenAI's GPT-6 Astra, leading on 12 of 18 disclosed benchmarks.
• Introductory pricing is set at $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 later, with cached inputs discounted 95%.
• The Fairwind Program, which opened September 3 with the smaller Gemini 3.8 Flash Cyber model, has already signed up more than 650 organizations including CrowdStrike and Palo Alto Networks.
• Google plans a guardrail-free version of Argon for trusted defenders and internal teams, and is participating in the U.S. government's voluntary pre-release model access process while hardening defenses against prompt injection and misuse.
• Shipping the model to security teams before the public shows AI labs increasingly treat vulnerability-hunting capability as both a defensive asset and a dual-use risk requiring staged, monitored release.