Gemini 4 Argon launch: Google limits new AI model to cyber defenders
Gemini 4 Argon launch: Google limits new AI model to cyber defenders
ISLAMABAD: Google on Thursday launched Gemini 4 Argon, its most powerful artificial intelligence model to date, but kept it out of public hands over safety concerns. The company said first access would go only to a small set of trusted cybersecurity defenders through its Fairwind Program.
The Gemini 4 Argon launch ends months of quiet development at Google. The company says the model is built to sustain deep reasoning across long, complicated workflows, and it claims frontier-level performance in software engineering, enterprise work such as legal and finance tasks, and cybersecurity defence.
Why the Gemini 4 Argon launch is restricted
Google said Argon’s capabilities are too sensitive for an open release. The model was trained specifically to find, validate and patch critical software vulnerabilities, a skill that could be used by criminals as well as defenders.
“Safely releasing frontier capabilities at this level requires a phased approach,” wrote Koray Kavukcuoglu, Google DeepMind’s senior vice-president and Google’s chief AI architect, in the announcement blog.
Google said it is taking part in the US government’s voluntary process for pre-release model access, and that it will collect feedback from early testers while it works on safeguards. The company has given no date for wider availability, saying only that developers, enterprises and consumers will get access as soon as possible.
The caution follows a White House meeting a day earlier, where President Donald Trump hosted tech executives including Google CEO Sundar Pichai and Anthropic’s Dario Amodei. The executives signed a voluntary accord pledging to police their own AI systems and disclose risks before widespread release.
What Argon can do
The model is designed for tasks that need sustained reasoning over many steps, rather than short answers. Its output limit has jumped to one million tokens, up from 64,000 in earlier Gemini models, letting it work through lengthy coding projects and business workflows in a single run.
Google shared internal examples of the model at work. Thousands of Google employees are already using Argon for debugging and codebase migrations, including a project that moved more than 800,000 lines of the Fuchsia Zircon kernel from C and C++ to the memory-safe Rust language. The company also claims Argon helped quantum-computing researchers improve an algorithm’s efficiency by 40 per cent over a published baseline, and freed more than 300 terabytes of memory by analysing data-centre telemetry.
On the security front, cybersecurity provider Wiz is one of the early users. Google said Wiz used Argon in its “Scan for Good” initiative to find and fix high-risk flaws in critical infrastructure, including a serious vulnerability in software used by hospitals.
Price and independent scores
Google set introductory API pricing at $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95 per cent.
Independent evaluator Vals Index gave Gemini 4 Argon 68.90 per cent in accuracy, placing it ahead of Anthropic’s Claude Sonnet 5.5, Claude Opus 5.5 and Claude Fable 5.1, according to Google’s announcement.
For Google, the launch is a bid to close the gap with rivals Anthropic and OpenAI in the race for frontier models. For everyone else, the wait continues: the company says the first expansion beyond cyber defenders will go to paid API customers and Google AI Ultra subscribers, with no timeline announced.
Frequently asked questions
When was Gemini 4 Argon announced?
Google announced the model on October 1, 2026, in an official blog post from Google DeepMind.
Who can use Gemini 4 Argon right now?
Only trusted cybersecurity defenders through Google’s Fairwind Program. Paid API customers and Google AI Ultra subscribers are expected to get access later, with no date given.
Why has Google restricted access to the model?
The company says Argon’s vulnerability-finding abilities carry safety risks. It is releasing the model in phases while it tests safeguards and takes part in the US government’s voluntary pre-release process.
How much does Gemini 4 Argon cost to use?
The introductory price is $2 per million input tokens and $10 per million output tokens, with cached inputs priced at 95 per cent off the input rate.
