Gemini 4 Argon Release Date: What We Know

There is no public Gemini 4 Argon release date yet. Google announced Argon on September 30, 2026, and so far it is only rolling out to a set of trusted cyber defenders through its Fairwind Program. Google says it will open access "as soon as possible," starting with paid Gemini API customers and Google AI Ultra subscribers, then developers, enterprises and consumers more broadly. It has not given a date for any of those steps.

Here is everything Google and independent testers have confirmed, what Argon will cost, and how to tell when it is actually coming to you. This page reflects the situation on October 5, 2026.

Gemini 4 Argon release date: the official position

In its launch post, Google DeepMind's Koray Kavukcuoglu said that "safely releasing frontier capabilities at this level requires a phased approach." Google says it is taking part in the US government's voluntary process for pre-release model access while it gradually widens availability, and that it will keep gathering feedback from early testers and refining guardrails before broader release.

The order Google described is:

1. Now: trusted cyber defenders in the Fairwind Program, plus Google's own internal teams. These testers get Argon without cyber guardrails so they can use its full vulnerability-finding ability.
2. Next: paid Gemini API customers and Google AI Ultra subscribers.
3. Then: developers, enterprises and consumers more generally.

As of today, Argon does not appear on the Gemini API models page, and Google has not published a public model ID. The Verge summed up the launch as Google saying the model is "so capable that only 'trusted cyber defenders' can have it right now" (The Verge).

Who can use Gemini 4 Argon today

Only Fairwind participants and Google employees. One public example is the security company Wiz, which Google says is using Argon in its free Scan for Good program for critical public infrastructure. If you are not a vetted cyber defender, you cannot use Argon in the Gemini app, Google AI Studio or the API yet.

Gemini 4 Argon pricing

Google has already published the API price, which is a useful signal that commercial launch is planned rather than open-ended.

In words: Argon will launch at two dollars per million input tokens and ten dollars per million output tokens, rising to four and twenty when the introductory period ends. Cached input tokens are 95 percent cheaper than regular input. Artificial Analysis says the discount lasts at least a month and that Google has not confirmed the end date.

For consumers, Argon is expected to arrive first in Google AI Ultra, which costs $99.99 or $199.99 a month in the US depending on usage level, per Google's subscriptions page. Google has not said whether Argon will later come to the $19.99 AI Pro plan.

What Argon can do, according to Google and testers

Google positions Argon as a model for long, complex workflows in software engineering, enterprise knowledge work such as legal and finance, and cybersecurity defense. Its headline claims:

• Output length: up to one million output tokens in a single response, up from 64,000, so the model can reason and write at enormous length.
• Coding: 77.9 percent on DeepSWE v1.1, which Google calls state of the art for long-horizon software engineering.
• Business work: first place on Zapier's AutomationBench at 51.3 percent, and the top score on the Vals Index of finance, coding, legal and tax work.
• Video: 91.7 percent on the LVBench long-video benchmark.
• Security: tied for first on CWE-bench v1 at 68 percent for fixing vulnerabilities.

Independent testing backs up the big picture. Artificial Analysis gives Argon a score of 53 on its Intelligence Index, level with GPT-6 Astra and one point ahead of GPT-6.1 Sol, and 23 points above Gemini 3.1 Pro Preview. It also measured the lowest hallucination rate among leading models on its knowledge test: 15 percent, compared with 51 percent for GPT-6 Astra. The trade-off is verbosity: Argon averaged about 62,000 output tokens per task against roughly 27,000 for Astra.

Pros: frontier-level intelligence, unusually low hallucination rate, very long outputs, aggressive introductory pricing. Cons: not available to most people, no firm date, uses many tokens per task, and price doubles after the intro period.

Why Google is holding Argon back

Google says it is strengthening four kinds of safeguards before a wide release: defenses against misuse for cyber or chemical, biological, radiological and nuclear attacks; resistance to prompt injection; monitoring of the model's chain of thought and actions for misalignment; and hardened, sealed sandboxes for high-risk training and evaluation.

That caution matches the rest of the industry this autumn. OpenAI restricts GPT-6 Astra's most advanced cyber abilities and cancelled a planned GPT-6.1 Astra over safety concerns, and Anthropic offers its least-restricted model, Mythos, only through trusted access programs. Our GPT-6 Astra vs Claude Fable 5.1 comparison and Claude Fable vs Opus vs Mythos guide explain those programs. Google DeepMind is the lab behind both Argon and Google's Frontier Safety Framework.

What about Gemini 3.5 Pro?

Gemini 3.5 Pro still has no release date either. In July, Google said 3.5 Pro was "currently testing with partners" and that it planned to release it "as soon as it's ready," while also revealing it had started its "most ambitious pre-training run yet, for Gemini 4" (Google's July model update). Since then Google has shipped several Flash models, most recently Gemini 3.8 Flash, but no new Pro.

That leaves Gemini 3.1 Pro as the top model most subscribers can use. It also matters for the October 9 Gemini app change: free users will be limited to Flash-Lite, AI Plus subscribers to Flash-Lite and Flash, and only AI Pro and Ultra keep the Pro model, according to 9to5Google. 9to5Google notes it is not yet clear whether Argon will slot in as a Pro model or as a new tier above it.

How to know when Argon is coming to you

Watch for these signals, roughly in this order:

1. A gemini-4-argon model ID on the Gemini API models and pricing pages.
2. Argon appearing in Google AI Studio's model picker for paid projects.
3. An announcement or help-center update naming Argon for Google AI Ultra.
4. A Gemini app model card or release note mentioning Argon for AI Pro.

Until then, if you need a frontier model today, compare the options in our ChatGPT vs Claude vs Gemini vs Grok guide. ChatGPT and Claude both offer their top models to paying subscribers now. For how Google's safety process compares with OpenAI's and Anthropic's, see Preparedness Framework vs Responsible Scaling Policy.

FAQ

When will Gemini 4 Argon be released to the public?

Google has not given a date. It says Argon will come to paid API customers and Google AI Ultra subscribers first, then to developers, enterprises and consumers, after it finishes strengthening safeguards with early testers.

How much will Gemini 4 Argon cost?

Through the API, $2 per million input tokens and $10 per million output tokens at introductory pricing, then $4 and $20. Cached input is 95 percent off. Consumers are expected to get it first through Google AI Ultra.

What is the Fairwind Program?

It is Google's trusted-access program for vetted cyber defenders. Fairwind participants are the first outside users of Gemini 4 Argon and receive it without cyber guardrails for defensive work.

Is Gemini 4 Argon better than GPT-6 Astra?

Artificial Analysis scores them level at 53 on its Intelligence Index, with Argon showing a much lower hallucination rate but using more tokens per task. Google's own benchmarks show Argon leading on several coding and business-workflow tests.

Is Gemini 3.5 Pro cancelled?

Google has not said so. Its last official word, in July, was that 3.5 Pro was testing with partners and would be released when ready.