Google has released Gemini 4 Argon, which it calls its most powerful model, at an introductory price of $2 per million input tokens and $10 per million output tokens. Almost nobody can buy it yet.
For now Argon is open only to members of Fairwind, a Google security programme that began on 2 September for governments, national cyber authorities, critical infrastructure operators and core technology platforms. They are meant to use it to find and fix holes in their own software before wider release.
Google says Argon can autonomously find, validate and patch critical vulnerabilities, as well as handle coding, long videos and multi-step work. In a post on Wednesday, Koray Kavukcuoglu, Google’s chief AI architect, said its output limit rises to one million tokens from 64,000.
On benchmarks, Google says Argon beats OpenAI’s GPT-6 Astra and Anthropic’s Fable and Opus models, and cites first place on the index run by benchmarking startup Vals. Those are Google’s numbers rather than independent tests.
The price is the more pointed claim. After the promotional period it doubles to $4 and $20 per million tokens, and it arrives barely a week after OpenAI and Anthropic cut their own prices within hours of each other.
Google can afford to stagger the launch. Its Gemini app passed one billion monthly users in August, matching ChatGPT, TechCrunch reported.
What Google has not given is a date. It says Argon will reach paid API customers and AI Ultra subscribers as soon as possible, which leaves a headline price attached to a model most developers cannot yet call.
