AI News · Models ·
Google announces Gemini 4 Argon with limited cybersecurity rollout
Google announced Gemini 4 Argon, claiming improvements in coding, cybersecurity and complex professional work, with a 1 million token limit. Access is initially restricted to trusted cyber defenders through its Fairwind Program while safety testing continues ahead of a wider release.
Key points
- Google is initially rolling out Gemini 4 Argon to trusted cyber defenders through its Fairwind Program.
- The model’s output limit rises to 1 million tokens, from 64,000.
- Introductory pricing is $2 per million input tokens and $10 per million output tokens.
- Google reports improvements across software engineering, professional workflows and defensive cybersecurity.
- Google has not disclosed a date for general availability.
What happened: Google announced Gemini 4 Argon, an AI model aimed at complex software engineering, legal and financial work, and cybersecurity defense. The company said access is initially rolling out to trusted cyber defenders through its Fairwind Program, rather than becoming broadly available to businesses or consumers. Google described the release as phased, with feedback from early testers informing further safeguards. It has not disclosed a date for general availability, making the announcement more relevant to evaluating future capabilities than planning an immediate deployment.
The details: Argon supports an output limit of 1 million tokens, up from the previous 64,000, according to Google. This is a limit on what the model can generate, not a stated expansion of how much information it can accept as input. Google said the additional capacity supports longer reasoning and multistep tasks. The company announced introductory pricing of $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95% from the input price.
Background: Google is positioning Argon around sustained professional workflows, including debugging, codebase migrations, financial research and legal drafting. The company reported a 77.9% score on DeepSWE v1.1, which evaluates extended software engineering tasks, and a first-place score of 51.3% on Zapier’s AutomationBench for end-to-end business execution. These are performance claims reported by Google. The company also said Argon is already used internally, including by agents working on migrations from C/C++ to Rust. Rewrites of critical systems are undergoing automated and manual auditing, testing and review before production rollout.
Who it affects: Cybersecurity defenders are the first external audience. Google said Argon can autonomously find, validate and patch critical software vulnerabilities, and that trusted defenders and its internal teams will receive access without cyber guardrails to use its full defensive capabilities. Wiz is already using the model through Scan for Good, its initiative to find and remediate high-risk exposures in critical public infrastructure for free. Google said an early deployment uncovered a critical vulnerability exposing sensitive personal information in healthcare software used by hospitals worldwide, which earlier models had missed.
What to watch: The central question for most business teams is when these capabilities will become accessible beyond the initial security rollout, and with what safeguards. Google said it is participating in the U.S. government’s voluntary process for pre-release model access and will continue testing before expanding availability to developers, enterprises and consumers. Its stated ambitions extend well beyond cybersecurity, but the restricted launch limits what most teams can evaluate today. Broader access and the safeguards accompanying it remain important details to establish before building deployment plans around Argon.
Our take
Argon is worth watching for complex workflows, but teams should not build deployment plans around it until broader access is confirmed.