# Gemini 4 Argon: Capabilities, Pricing and Access

By Schnalz.net · 1 October 2026

https://schnalz.net/ai-guides/gemini-4-argon-capabilities-pricing-and-access

> Gemini 4 Argon is Google's frontier AI model designed for long-horizon tasks across coding, cybersecurity, and enterprise work. It features a 1 million output token capacity and is currently rolling out to early testers through the Fairwind Program.

## In short

- Google announced Gemini 4 Argon on 30 September 2026.
- The model features an output token limit of 1 million tokens.
- Introductory pricing is set at $2 per million input tokens and $10 per million output tokens.
- Argon is currently accessible only to select cyber defenders via the Fairwind Program.

## What is Gemini 4 Argon?

Gemini 4 Argon is a frontier artificial intelligence model developed by Google and Google DeepMind for complex, long-horizon workflows. Announced on 30 September 2026, the model focuses on software engineering, enterprise knowledge work, and cybersecurity defense.

**Gemini 4 Argon details at a glance**

- **Developer**: Google DeepMind
- **Announcement date**: 30 September 2026
- **Output token limit**: 1 million tokens
- **Input token price**: $2 per million tokens
- **Output token price**: $10 per million tokens
- **Current access status**: Fairwind Program early testers

Google uses Gemini 4 Argon internally for daily engineering tasks, codebase migrations, and data center optimizations. The system supports multimodal analysis, allowing it to process documents, evaluate visual charts, and analyze long videos.

## Gemini 4 Argon Key Capabilities and Benchmarks

Gemini 4 Argon introduces an output token capacity of 1 million tokens, expanding from the previous 64,000 token limit. This expanded headroom allows Gemini 4 Argon to sustain long reasoning trajectories and solve multi-step problems in a single pass.

- **1M** Output token limit
- **77.9%** DeepSWE v1.1 benchmark score
- **91.7%** LVBench long video score
- **51.3%** AutomationBench score (#1 rank)

According to Google, Gemini 4 Argon holds top scores across multiple industry evaluations. It scored 77.9% on the DeepSWE v1.1 benchmark for software engineering and achieved a state-of-the-art score of 91.7% on LVBench for video understanding.

The model leads the Vals Index across finance, coding, legal, and tax evaluations. On Zapier's AutomationBench, which tests core business task execution, Gemini 4 Argon ranks first with a score of 51.3%.

## Cybersecurity Capabilities and the Fairwind Program

Google specifically trained Gemini 4 Argon to assist in defensive cybersecurity workflows. Gemini 4 Argon can autonomously discover, validate, and patch critical vulnerabilities across codebases in 20 programming languages.

On CWE-bench v1, which evaluates security patch creation, Gemini 4 Argon tied for first place with a 68% score. Google provides versions of Gemini 4 Argon without cyber guardrails to trusted defenders so they can access full defensive features.

Cybersecurity company Wiz uses Gemini 4 Argon in its Scan for Good initiative to protect public infrastructure. In early tests, Gemini 4 Argon discovered a critical exposure in healthcare software that earlier AI models had missed.

## Gemini 4 Argon Pricing and Token Limits

Google announced introductory pricing for Gemini 4 Argon ahead of its general availability release. Standard input usage costs $2 per million tokens, while generated outputs cost $10 per million tokens.

Google offers a 95% discount on cached input tokens compared to standard input token pricing. The expanded 1 million token output limit allows developers and enterprises to generate hundreds of thousands of output tokens in continuous runs.

## How to Access and Use Gemini 4 Argon

Gemini 4 Argon is not yet available for public sign-up, self-service developer access, or general commercial deployment. Google is currently limiting access to selected cyber security partners through the Fairwind Program.

> **Note**: Gemini 4 Argon cannot be accessed by the general public yet. Google is gathering feedback from Fairwind Program participants and testing guardrails before opening access to developers and enterprise customers.

Google stated that it is engaged in the voluntary pre-release access process established by the U.S. government. The company plans to expand access to developers, enterprise organizations, and consumers once testing and safety reviews finish.

## Frequently asked questions

### What is Gemini 4 Argon?

Gemini 4 Argon is Google's frontier AI model designed for long-horizon problem solving in software engineering, cybersecurity, finance, and legal work. Google announced the model on 30 September 2026.

### How much does Gemini 4 Argon cost?

Google set introductory pricing for Gemini 4 Argon at $2 per million input tokens and $10 per million output tokens. Cached input tokens receive a 95% discount off standard input token rates.

### Is Gemini 4 Argon available to the public?

No, Gemini 4 Argon is not currently available to the public or standard developers. Google is deploying the model selectively to cyber defenders through its Fairwind Program before a wider public rollout.

### What is the token limit of Gemini 4 Argon?

Gemini 4 Argon supports an output token limit of up to 1 million tokens. This represents an increase from the previous 64,000 token output limit in prior Google models.

## Sources

- [Google: Introducing Gemini 4](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/)