Google launches Gemini 4 Argon, an AI model for complex professional tasks
1–2 minutes
On September 30, Google announced its new Gemini 4 Argon model, initially available only to select cybersecurity partners in the Fairwind program. The system is designed to support multi-step professional workflows, such as programming, financial research, legal document drafting, and software vulnerability detection.
Who will be able to use it?
After the pilot phase, the company plans to expand access to developers, businesses, and consumers, starting with paying API customers and Google AI Ultra subscribers. No specific date has been set for this expansion yet.
The introductory API pricing is $2 per million input tokens and $10 per million output tokens. Cached and reused content receives a 95% discount on the standard input rate.
What does 1 million tokens mean?
The model’s generation limit has been increased from 64,000 to 1 million output tokens. Tokens represent units of information processed by the AI, which can correspond to words or parts of words in text.
This limit refers to the amount of text the model can generate in a single run, not the volume of documents it can receive as input. The increase enables more extensive reasoning and results, although not all responses reach the maximum limit.
Why security comes first
Initial access is prioritized for organizations specializing in cybersecurity. Google claims that Argon can identify, validate, and fix vulnerabilities, and that select partners will have access without the usual restrictions in this field.
Before a wider release, the company is strengthening protections against misuse, malicious instructions in external content, and actions that exceed user intent. Feedback from early participants will be used to refine these controls.
Frequently Asked Questions
Is it already possible to use Gemini 4 Argon?
Initial access is limited to select cybersecurity partners. There are currently no plans for widespread distribution.
How much will the model cost?
The introductory API price is $2 per million input tokens and $10 per million output tokens, with a 95% discount for reused cache inputs.
Is the 1 million token limit for documents sent?
No. The announced limit refers to the model’s output generation, which is designed for more extensive runs and results.
