Google Gemini 4 Argon: Features, 1 Million Token Output and Availability

• Google Gemini 4 Argon targets complex software engineering and long-horizon workflows.
• The model can generate up to one million output tokens for extended tasks.
• Initial access is limited to trusted cyber defenders and selected testers.

Google Gemini 4 Argon is the first model announced in Google’s Gemini 4 series, with a focus on complex, long-running tasks rather than simple chatbot queries. Google says the model is designed for software engineering, professional knowledge work and cybersecurity defence.

Google announced Gemini 4 Argon on September 30, 2026, but it is not yet generally available. The initial rollout is going to trusted cyber defenders and selected testers, with paid API customers and Google AI Ultra subscribers planned for a later stage.

Google Gemini 4 Argon Features

Gemini 4 Argon is built for tasks that require multiple connected steps, including coding, analysis and professional workflows. Google has also highlighted multimodal capabilities and long-form generation as part of the model’s expanded capabilities. For more AI coverage, see AI news and updates on MobileTelco.

• Long-horizon reasoning for complex, multi-step workflows.
• Advanced software engineering and coding capabilities.
• Professional knowledge work across areas such as finance and legal tasks.
• Cybersecurity vulnerability analysis and defensive security work.
• Multimodal understanding, including long-video analysis.

Gemini 4 Argon 1 Million Token Output

One of the major changes with Gemini 4 Argon is its maximum output limit. Google says the model can generate up to 1 million output tokens, compared with a 64K output limit on earlier Gemini models.

The larger output capacity is intended for extended reasoning and long-running tasks. It should not be interpreted as a one-million-token input or context-window specification, as those are separate limits.

Gemini 4 Argon Coding and AI Workflows

Google says Argon is designed to handle software engineering work that can require planning, coding, debugging and repeated iterations. The company has also described internal uses involving code migration, data-centre memory optimisation and software performance work. Google’s wider app ecosystem also includes features covered in our Google Play custom game playlists guide.

Google reports a 77.9% result on the DeepSWE v1.1 software-engineering benchmark and 91.7% on LVBench for long-video understanding. These are Google’s reported benchmark results, so they should be treated as vendor-reported measurements rather than independent testing.

Gemini 4 Argon for Cybersecurity

Cybersecurity is a major part of the initial Gemini 4 Argon rollout. Google is first making the model available to trusted cyber defenders through its Fairwind programme, allowing selected users to evaluate its capabilities in security-related work.

Google has also described safeguards and monitoring designed around the risks associated with highly capable AI systems. The controlled rollout reflects the company’s decision to test the model before wider availability.

Gemini 4 Argon Availability

Gemini 4 Argon is currently available only to a limited group of trusted cyber defenders and testers. Google says paid API customers and Google AI Ultra subscribers are next in its rollout plan, but it has not announced a specific date for wider access.

Gemini 4 Argon Price

For API customers, Google has announced introductory pricing of $2 per million input tokens and $10 per million output tokens. Cached input is offered at a substantially lower rate. Google says the pricing is introductory, so developers should check the latest API pricing before deployment.

Gemini 4 Argon Benchmarks

Benchmark	Google-reported result
DeepSWE v1.1	77.9%
LVBench	91.7%
CWE-bench v1	68%
AutomationBench	51.3%

Benchmark scores use different tests, datasets and evaluation methods, so the figures should not be treated as directly interchangeable. The results above are based on Google’s published claims and are not independent MobileTelco tests.

Gemini 4 Argon vs Gemini 3

Gemini 4 Argon changes the focus from shorter model interactions towards sustained, multi-step work. Its 1-million-token output ceiling, coding focus and controlled cybersecurity rollout are among the clearest differences highlighted by Google. The company has not positioned the announcement as a simple replacement with identical access across existing Gemini products.

Conclusion

Google Gemini 4 Argon is a new frontier model focused on complex coding, professional workflows, multimodal tasks and cybersecurity. Its 1-million-token output limit is a major technical change, while access remains restricted during the initial rollout. Wider availability for paid API customers and Google AI Ultra subscribers is planned, but Google has not provided a public release date.

Frequently Asked Questions