Gemini 4 Argon: Release Date, Access, Pricing and What Enterprises Should Do While They Wait

Google has a new frontier model, and almost nobody can use it yet.

On September 30, Google announced Gemini 4 Argon. It is rolling out first to a set of trusted cyber defenders through Google’s Fairwind Program. Developers, enterprises and consumers come later. If you are asking when you can use it, what it costs and whether to plan around it, this post collects what Google has actually said.

Gemini 4 Argon Release Date and Access

There is no general release date. Google says it is taking a phased approach: Argon is going to trusted cyber defenders first, while Google takes part in the U.S. government’s voluntary process for pre-release model access and gradually expands availability. It says it will make Argon available to developers, enterprises and consumers “as soon as possible,” with no date attached.

Axios reports that Google says paying subscribers will be first in line when access expands. The Verge adds that Google will strengthen safeguards against misuse and prompt injection, and monitoring for misalignment, before a broader rollout.

This is our planning advice, not a date from Google: assume Argon is not available to your team this quarter unless you are a cyber defender in the program, and treat any date you see online as speculation.

Gemini 4 Argon Pricing

Google lists an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens at 95% off the input price. Google’s footnote says that after the introductory period expires, the price becomes $4 per million input tokens and $20 per million output tokens. Google does not say in that post when the introductory period ends, so budget with the $4 and $20 figures, and note that you cannot buy access at either price yet unless you are in the program.

What Google Says Argon Can Do

Every claim in this section is Google’s own, from its launch post. These are vendor benchmarks and have not been independently reproduced.

• Software engineering. Google reports a state-of-the-art 77.9% on DeepSWE v1.1, a benchmark of long-horizon software tasks.
• Enterprise knowledge work. Google says Argon leads the Vals Index, which measures economic impact across finance, coding, legal and tax work, and is state of the art on Vals Finance Agent v2 and Harvey’s Legal Agent Benchmark.
• Business automation. On Zapier’s AutomationBench, which measures end-to-end execution across core business functions, Google reports a #1 score of 51.3%.
• Long output. The output token limit rises to 1M tokens, up from 64K, so the model can reason and write across very long single tasks.
• Cybersecurity. Google says Argon can find, validate and patch critical vulnerabilities, and ties for first on CWE-bench v1 at 68%.
• Internal use. Google describes Argon agents migrating C/C++ code to Rust, including a libgav1 video decoder it says runs 2.7x faster than the earlier Rust port, and finding memory optimizations in its data centers.

Gemini 4 Argon vs GPT-6 Astra

Axios reports that Google says Argon outperforms OpenAI’s GPT-6 Astra on several coding and knowledge-work benchmarks. We have not found independent head-to-head testing yet, so treat this as a claim. There is also a context point: OpenAI said this week it would not release a planned GPT-6.1 Astra model over safety concerns, according to The Verge and Axios. Both vendors are now publicly slowing releases on safety grounds.

What Enterprises Should Do While They Wait

You cannot test Argon, but you can get ready for it. A better model fails differently, not less often in the places that matter to you.

1. Build your evaluation set now. Collect 50 to 100 real tasks from your own workflows, with the answer you would accept. When any new model becomes available, you can score it in a day instead of arguing about benchmarks. This is the single most useful preparation, and it is part of how we approach agentic AI systems.
2. Make your stack model-agnostic. If your prompts, tools and logging are hard-wired to one vendor, switching costs you weeks. Keep a thin abstraction layer and test two models side by side. See our notes on agent orchestration.
3. Budget for output-heavy workloads. The pricing listed for Argon is five times higher for output than input. Long-running agents that write a lot will cost more than their prompts suggest.
4. Set a security gate. Google’s own launch stresses prompt-injection resistance and misalignment monitoring. Ask the same questions of every agent you deploy, as we argued in OpenAI Dots vs Meta Muse.
5. Check your regulatory exposure. If you operate in the EU, the obligations of the AI Act apply regardless of which model you pick: EU AI Act compliance is an engineering problem.

The same week, Meta pushed into enterprise AI too. We cover what that means for customer messaging in Meta Enterprise Platform and WhatsApp Business automation.

If you want help turning this into a working evaluation harness and a rollout plan, our AI strategy team does exactly that. Talk to us.

Frequently Asked Questions

When will Gemini 4 Argon be available?

Google has not given a date. It is available first to trusted cyber defenders through the Fairwind Program, and Google says it will widen access to developers, enterprises and consumers as soon as possible after more safeguard work.

How much does Gemini 4 Argon cost?

Google lists an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens 95% cheaper than input. According to Google’s footnote, the price after the introductory period is $4 per million input tokens and $20 per million output tokens. Google’s post does not give the end date of the introductory period.

Can I use Gemini 4 Argon today?

Only if you are one of the trusted cyber defenders Google has selected. Everyone else is waiting.

Is Gemini 4 Argon better than GPT-6 Astra?

Google says it beats GPT-6 Astra on several benchmarks, per Axios. Those are Google’s results, and independent comparisons are not available yet.

Why is Google limiting access to Gemini 4 Argon?

Google says safely releasing a model at this capability level needs a phased approach. It is working through the U.S. government’s voluntary pre-release process and strengthening protections against misuse, prompt injection and misalignment.

The Bottom Line

Gemini 4 Argon is a signal more than a tool for now: Google is back at the top of the benchmark tables, and frontier labs are releasing in stages. The practical move is not to wait for access. Build the evaluation set, keep your stack portable and set your security gate, so that whichever model arrives first, you can judge it in days.