TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Google announced Gemini 4 Argon on Sept. 30, 2026, and said it is beginning a limited rollout to trusted cyber defenders through its Fairwind Program. Google reports strong results in software engineering, enterprise tasks and defensive cybersecurity, but those performance claims are company-reported and broad public access has not begun.
Google announced Gemini 4 Argon on Sept. 30, saying it is beginning a limited release to trusted cyber defenders through its Fairwind Program. The company describes Argon as a model for long, complex work in software engineering, legal and finance tasks, and cybersecurity defense; it has not yet opened the model to developers, businesses or consumers generally.
Google says Argon can sustain reasoning across complex workflows and has already been used by thousands of its employees for coding, research and writing. The company reports that the model set a benchmark high on DeepSWE v1.1, a test of long-horizon software engineering, with a score of 77.9%. Google also reports a score of 51.3% on AutomationBench, where it ranked first, and 91.7% on LVBench, a long-video understanding evaluation. These are results disclosed by Google; the announcement does not provide independent verification or full benchmark methods.
The company described internal engineering projects as examples of Argon’s use. It said agents helped researchers improve a quantum-computing subroutine by 40% against a published baseline, and identified memory changes that could free more than 300 tebibytes once deployed, with estimated total savings of 500 tebibytes to one pebibyte. Google also says Argon agents are assisting with C and C++ to Rust migrations, including work on codebases as large as the Fuchsia Zircon kernel. It says those projects undergo automated and manual audits, emulation testing and review before production use.
For the open-source video decoder libgav1, Google says agents replaced 32,000 lines of SIMD code in an existing Rust port through repeated profiling and compiler experiments. The company reports that the resulting decoder runs 2.7 times faster than that Rust port while producing identical video output. It says the work brings performance closer to the optimized C++ version. Google has not supplied independent measurements in the announcement.
A Cautious Route to Powerful Tools
Argon’s initial distribution places cybersecurity defense at the center of its launch. Google says the model can autonomously find, validate and patch critical software vulnerabilities, and says trusted defenders and its internal teams will receive it without cyber guardrails to use its full defensive capabilities. That access could help security teams investigate and fix flaws, while the same capabilities make staged access and safeguards a central part of the release.
The announcement also signals an attempt to make AI useful across work that involves more than short coding prompts. Google says Argon can process tasks requiring code, documents, financial research and visual information. Its expanded one-million-token output limit, up from 64,000 tokens in the previous limit cited by the company, is intended to let the model generate much longer responses in a single trajectory. The practical value of that capacity will depend on reliability, cost and how well the model handles extended tasks outside Google’s own evaluations.
Google lists an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens priced 95% below the input rate. These prices provide an early reference for developers and enterprises evaluating potential use, but broad availability has not started. The staged rollout means interested customers cannot yet assume they can access or deploy Argon.
From Internal Use to Early Access
Google presents Argon as an extension of its Gemini model work, focused on tasks that require sustained reasoning and multiple steps. The company says the model is already being used internally for engineering productivity, research and writing. Its announcement highlights work on algorithm design, large code migrations and analysis of company-wide profiling data as examples of that internal use.
For external access, Google says it is participating in the U.S. government’s voluntary process for pre-release model access while gradually expanding testing. The first announced group is trusted cyber defenders in the Fairwind Program. Google says it will gather feedback from early testers and adjust safeguards before making Argon available to developers, enterprises and consumers. The announcement does not give a schedule for those later stages.
““Safely releasing frontier capabilities at this level requires a phased approach.””
Questions Before Wider Release
Google has not said when Argon will become available to developers, enterprises or consumers, how large the Fairwind rollout will be, or which testers have access beyond its description of trusted cyber defenders. Details of the government pre-release process and the criteria for moving to broader access are also not included in the announcement.
The performance figures and project outcomes cited so far come from Google. The announcement does not provide independent evaluations, full test conditions for each benchmark, or enough detail to assess how results compare across models under consistent conditions. It also does not report how often Argon’s vulnerability findings are valid, how reliably its proposed patches work, or what review is required before those changes are used. Those questions matter for judging real-world value and risk.
Tester Feedback Sets the Pace
Google says it will expand access gradually, collect feedback from early testers and revise safeguards before a wider release. The next visible milestones are likely to be further details about the Fairwind testing group, evidence from external evaluations and a schedule for access beyond trusted defenders. The company has not announced dates for those steps.
When access broadens, Google’s published introductory rates offer an initial basis for estimating usage costs. For now, the model remains in a limited rollout, and the timing and terms of general availability are still open.
Key Questions
What is Gemini 4 Argon?
Google describes Argon as a model for complex, long-running work in software engineering, enterprise knowledge tasks and cybersecurity defense.
Who can use Argon now?
Google says it is beginning a rollout to trusted cyber defenders through its Fairwind Program. General access for developers, enterprises and consumers has not started.
What results has Google reported?
Google reports scores of 77.9% on DeepSWE v1.1, 51.3% on AutomationBench and 91.7% on LVBench, alongside internal engineering examples. These are company-reported results; the announcement does not provide independent verification.
When will Gemini 4 Argon be broadly available?
Google says it plans to expand access after gathering tester feedback and updating safeguards, but has not announced a date.
Source: hn
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
