Free tools Windows power users keep installed
One-click scans. No signup required.
Google announced Gemini 4 Argon on September 30, 2026, positioning it for complex software engineering, enterprise work and cybersecurity defense. The million-token figure is an output limit, not a published context-window size: Google has not stated Argon’s input limit or total context window.
What is Gemini 4 Argon designed to do?
Google describes Argon as a frontier model for long-horizon software engineering, enterprise knowledge work such as finance and legal tasks, and cybersecurity defense. The company says it can find, validate and patch critical software vulnerabilities. Those are Google’s capability claims, not results independently established here.
As an Amazon Associate I earn from qualifying purchases.
Google also reports that its Scan for Good initiative is using Argon for defense through Wiz, and says the work included discovery of a critical vulnerability affecting healthcare software. The announcement does not name the software or provide further details, so the claim cannot be assessed more specifically from that account.
Is the 1 million-token figure a context window or an output limit?
It is an output-token limit. In its September 30 announcement, Google said Argon’s maximum output had increased from 64,000 tokens to 1 million, to support longer reasoning and generation trajectories. Google’s October 2 monthly roundup also calls it a 1-million-token output limit.
#1 Best Overall
Neither announcement specifies Argon’s input-token limit or its total context window. Those figures describe different capabilities: output is what a model can generate, while a context window concerns the information it can take into account across input and output. Argon should not be described as having a 1-million-token context window based on the published figure.
What benchmark results has Google reported?
The figures below are reported by Google in its September 30 announcement. The company describes the rankings, but the announcement does not provide independent replication or enough detail to treat the scores as a prediction of performance on every real-world task.
| Evaluation | Google-reported result | What Google says it measures |
|---|---|---|
| DeepSWE v1.1 | 77.9% | Real-world, long-horizon software-engineering tasks |
| AutomationBench | 51.3%; Google says Argon ranks first | Automation tasks |
| LVBench | 91.7%; Google describes the result as state of the art | Long-video understanding |
| CWE-bench v1 | 68%; Google says Argon ties for first | Vulnerability remediation |
Google also says Argon leads the Vals Index and reports strong results on Vals Finance Agent v2 and Harvey’s Legal Agent Benchmark, without giving numeric scores in the announcement. It says Wiz’s internal black-box penetration-testing benchmark put Argon ahead of Gemini 3.8 Flash Cyber for attack-surface discovery, vulnerability identification and proof-of-concept validation; no score is given.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsWhat can Argon do for cybersecurity, and what safeguards does Google describe?
Google presents Argon as a defensive tool for finding, verifying and fixing vulnerabilities, including through autonomous work. Its account describes safeguards intended to manage the risks of capable cyber tools:
Rank #3
- Refusing harmful cyber and chemical, biological, radiological and nuclear requests while allowing legitimate dual-use work.
- Testing safeguards with internal and external red teams.
- Training and testing against indirect prompt injection, which can attempt to steer a model through instructions embedded in content it processes.
- Monitoring chain-of-thought and actions to stop behavior that goes beyond user intent.
- Isolating sandbox environments before high-risk training or evaluation.
Google says trusted defenders and its internal teams will receive Argon without cyber guardrails. This describes Google’s stated deployment approach; it is not independent certification that the model or safeguards are safe in every setting. As Google DeepMind SVP and Google Chief AI Architect Koray Kavukcuoglu put it, “Safely releasing frontier capabilities at this level requires a phased approach.”
Who can use Gemini 4 Argon, and is it generally available?
Google says access is starting with a set of trusted cyber defenders through its Fairwind Program. The company plans to use feedback to iterate on guardrails before expanding access to developers, enterprises and consumers. It names paid API customers and Google AI Ultra subscribers as starting audiences for broader release, but gives no general-availability date.
Google’s September 2 description of Fairwind characterizes it as limited access for governments, trusted Google Cloud customers and cybersecurity partners. Participating organizations are expected to restrict access to internal cybersecurity, incident-response or penetration-testing teams and use protections such as multifactor authentication. That program description does not mean every Fairwind participant has Argon access.
Recommended Free Tools
How much does Gemini 4 Argon cost?
Google announced these API rates. The announcement does not say when billing starts or how long the introductory period lasts.
Best Value
| Token type | Introductory announced rate | Rate after introductory period |
|---|---|---|
| Input | $2 per million tokens | $4 per million tokens |
| Output | $10 per million tokens | $20 per million tokens |
| Cached input | 95% below the introductory input rate | Not stated in the announcement |
These are announced API prices, not a confirmed billing start date or a guarantee that access is currently available to every API customer.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
