Gemini 4 Argon
AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

Prime Big Deal Days · Oct 6–7Offer from Amazon

Get tech for your team delivered free — and shop member deals

  • Fast, free delivery on millions of items
  • Access to Prime Big Deal Days deals on October 6–7
  • Prime Video, Amazon Music and more included
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

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 describes the model as capable across software engineering, enterprise work and cybersecurity, but its performance figures and internal productivity claims come from the company and cited benchmark providers.

Google announced Gemini 4 Argon on Sept. 30, saying the model is beginning a limited release to trusted cyber defenders through its Fairwind Program. The company says Argon is designed for long, complex tasks in software engineering, enterprise work and cybersecurity; access will expand gradually as Google collects tester feedback and revises safeguards.

Google says Argon can sustain reasoning across long workflows and is already being used internally by thousands of employees for coding, research and writing. The company reports that one quantum-computing task beat a published baseline by 40%. It also says an internal agent project identified memory optimizations that, once rolled out, freed more than 300 TiB across its data centers, with estimated total savings of 500 TiB to 1 PiB. Those figures are Google’s reported results; the announcement does not provide independent evaluations of the projects.

For software migration, Google says Argon agents are helping move C and C++ code to Rust, including work on codebases ranging from tens of thousands of lines to more than 800,000 lines in the Fuchsia Zircon kernel. The company says the migrations are subject to automated and manual audits, emulation tests and review before production use. In a separate example, Google reports that an Argon-assisted rewrite replaced 32,000 lines of SIMD code in a Rust version of its libgav1 video decoder. It says the result ran 2.7 times faster than the earlier Rust port while producing identical video output.

Google also highlights third-party benchmarks: it reports a 77.9% score on DeepSWE v1.1, a ranking at the top of the Vals Index, and a first-place score of 51.3% on Zapier’s AutomationBench. The company says Argon scored 91.7% on LVBench for long-video understanding. These are benchmark results cited in Google’s announcement, not evidence by themselves of performance in every workplace or real-world deployment. Google says the model supports up to 1 million output tokens, compared with 64,000 for the prior limit it references. Its announced introductory prices are $2 per million input tokens and $10 per million output tokens; cached input tokens are priced 95% below the input rate.

At a glance
announcementWhen: Announced Sept. 30, 2026; limited rollo…
The developmentGoogle announced Gemini 4 Argon, a frontier model initially being made available to selected cyber defenders while the company gathers feedback and updates safeguards.

A Limited Release for Cyber Defense

The rollout makes access and safeguards central to the announcement. Google says trusted defenders will receive Argon without cyber guardrails so they can use its cybersecurity defense capabilities, including finding, validating and patching software vulnerabilities. The company says it is participating in a voluntary U.S. government process for pre-release model access and will widen availability as it gathers feedback and iterates on protections.

If the reported capabilities hold up in use, a model that can carry out extended coding and research workflows could affect how organizations handle software maintenance and knowledge work. The cyber focus also gives defenders a potential tool for remediation, while making decisions about access and oversight consequential. The current announcement describes a controlled initial audience; it does not establish how broadly the model will be deployed or how well its capabilities generalize outside Google’s examples and benchmark tests.

From Internal Use to Early Access

Google presents Argon as both an internal productivity model and a system intended for external use. Its announcement describes ongoing internal projects, including code migrations and data-center memory work, while also setting out a staged release. The first named external pathway is the Fairwind Program for trusted cyber defenders. Google says broader access for developers, enterprises and consumers is planned as soon as possible, following further testing and changes to safeguards.

The announcement names Wiz as an early user of Argon for cybersecurity defense through its Scan for Good initiative. It does not give further details in the material provided about the initiative’s results, the specific access terms, or which other organizations are participating. Google’s benchmark claims span software engineering, finance, legal work, business automation and video analysis, but each measure covers a particular test and should be read within that scope.

““Safely releasing frontier capabilities at this level requires a phased approach.””

— Google

Limits of the Early Evidence

The announcement does not state when Argon will become generally available or identify a date for each rollout stage. It also does not specify the full evaluation criteria behind every benchmark claim or provide independent validation of the internal project results. The expected memory savings are estimates, and Google says the code migrations require audits and review before production rollout.

Details of the Fairwind participants, the practical scope of access without cyber guardrails, and the safeguards planned for later users are not fully described in the available announcement. Google says the U.S. government process is voluntary, but the material does not explain its procedures or outcome. The claims about vulnerability discovery and patching describe intended capability; the announcement supplies no case-by-case results showing how Argon performs against deployed threats.

Feedback Before Wider Availability

Google says it will collect feedback from early testers, continue refining safeguards and gradually expand access. The company has not set a public launch date for developers, enterprise customers or consumers. Further information about rollout timing, access conditions and independent evaluations would clarify how the initial results translate to broader use.

Key Questions

What is Gemini 4 Argon?

Google describes Argon as a frontier model for extended tasks in software engineering, enterprise knowledge work and cybersecurity defense. The company says it supports up to 1 million output tokens.

Who can use Argon now?

Google says it is beginning a limited rollout to trusted cyber defenders through its Fairwind Program. It has not published a full participant list or general access date in the announcement.

When will Argon be available to the public?

Google says it plans to make Argon available to developers, enterprises and consumers after gathering feedback and refining safeguards, but it has not announced a specific date.

What price did Google announce?

The introductory price is $2 per million input tokens and $10 per million output tokens. Google says cached input tokens cost 95% less than the standard input rate.

Are the performance claims independently verified?

The announcement cites benchmark scores and reports results from Google’s internal projects, but the supplied material does not establish independent verification of those claims. Benchmark scores describe performance on particular tests and do not by themselves predict results in every deployment.

Source: hn

FALL

Fall Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

What Society Should Know About Anthropic’s New AI Watermarking System

Anthropic has launched a new watermarking system for its Claude AI, aiming to improve content provenance verification amid ongoing technical uncertainties.

Improve Your AI Search Ranking With ChatGPT’s Monitoring Solutions

New ChatGPT rank monitor helps brands track AI visibility, share-of-voice, and citations, addressing a growing need amid shifting search behaviors.

Is LFM2.5-DSpark The Key To Faster AI Inference? Discover The 3.2X Speed Boost

LiquidAI releases DSpark draft checkpoints for LFM2.5 models, delivering up to 3.18x GPU speedup and 2.87x on-device without quality loss, revolutionizing AI inference.

The Meticulous AI That Couldn’t Finish the Job

Firmulate’s most diligent AI added 80 rules and found every crisis, yet finished last—showing why disciplined execution matters more than volume.