# Gemini 4 Argon Release Date: Is It Out, and Who Has Access?
By SeedRouter · Published 2026-10-01 · Updated 2026-10-01
**Google announced Gemini 4 Argon on September 30, 2026, but it is not out for most people yet.** It is rolling out first to a set of trusted cyber defenders through Google's Fairwind Program. Developers, enterprises and consumers come later, and Google has not given a date.\[1]
So there are two answers to "what is the Gemini 4 Argon release date". The announcement date is September 30, 2026. The date you can actually call it from your own code has not been set.
## Is Gemini 4 Argon out?
Only for a small group. Google describes a phased release while it takes part in the U.S. government's voluntary process for pre-release model access and keeps working on safeguards with early testers.\[1] The order Google has published is:
| Phase | Who | Status on October 1, 2026 |
| ------------------- | -------------------------------------------------- | ------------------------- |
| Internal use | Google's own teams | In use |
| Early access | Trusted cyber defenders in the Fairwind Program | Rolling out |
| First wider release | Paid API customers and Google AI Ultra subscribers | No date announced |
| Broad availability | Developers, enterprises and consumers | No date announced |
If you are not a Fairwind partner, there is no way to use Gemini 4 Argon today. That includes the public Gemini API: on October 1, 2026, Google's model list for the Gemini API did not include Argon, and no API model ID had been published.\[4]
## Who can use Gemini 4 Argon today?
The Fairwind Program is Google DeepMind's access program for approved cyber defense organizations. A set of its partners get exclusive access to Gemini 4 Argon, either on its own or inside CodeMender, Google's code security agent, to find and patch vulnerabilities.\[2]
Access comes with conditions. Partner organizations agree to user-level authentication, phishing-resistant MFA and access controls. They may give access only to internal cybersecurity, incident response or penetration testing teams, and must track who uses it.\[2] For these defenders and its own teams, Google is releasing Argon without its cyber guardrails; the wider release keeps safeguards that refuse harmful cyber and CBRN requests.\[1]
## When will developers get Gemini 4 Argon?
Google says the wider release will start with paid API customers and Google AI Ultra subscribers, "as soon as possible", after it has iterated on guardrails with early testers.\[1] That is the whole commitment. There is no month, no region list and no waitlist in the announcement.
Two things are already fixed, though: the price and the output limit.
## What will Gemini 4 Argon cost?
Google published API prices in the launch post, even though the API is not open yet:\[1]
| Token type | Introductory price | After the introductory period |
| ------------ | -------------------------------------- | ----------------------------- |
| Input | $2 per 1M tokens | $4 per 1M tokens |
| Output | $10 per 1M tokens | $20 per 1M tokens |
| Cached input | 95% off the input price ($0.10 per 1M) | Not announced |
Prices as published by Google, checked on October 1, 2026.
Google has not said how long the introductory period lasts. Its output limit is 1M tokens, up from the previous 64K, so a single long response can be expensive at $10 per 1M output tokens: one full-length answer is up to $10 of output on its own.\[1]
## How does Gemini 4 Argon compare at launch?
Google published 18 benchmarks against GPT-6 Astra, Claude Fable 5.1 and Claude Opus 5.5. By our count of that table, Argon leads on 12, ties GPT-6 Astra on CWE-bench v1, and trails on five.\[3]
| Benchmark | Gemini 4 Argon | Best other model in Google's table |
| ----------------------------------- | -------------- | ---------------------------------- |
| DeepSWE v1.1 (agentic coding) | 77.9% | 74.2% (Claude Opus 5.5) |
| Vals Index (knowledge work) | 68.9% | 67.0% (Claude Opus 5.5) |
| LVBench (long video) | 91.7% | 87.5% (GPT-6 Astra) |
| CWE-bench v1 (cybersecurity) | 68.0% | 68.0% (GPT-6 Astra) |
| FrontierSWE v2 (agentic coding) | 55.0% | 65.5% (GPT-6 Astra) |
| Terminal-bench 4.0 (agentic coding) | 57.4% | 66.4% (Claude Opus 5.5) |
These are Google's numbers. Argon was run at its highest thinking setting, and most results for the other models are their developers' own reported scores.\[5] The full table is on the [Gemini 4 Argon model page](https://seedrouter.ai/models/gemini-4-argon).
## What has Google not announced?
* **An API model ID.** No ID appears in the public Gemini API model list yet.\[4]
* **The input context window.** Google states the 1M-token output limit, not the input limit. Its long-context results include inputs of up to 1M tokens.\[3]
* **Dates.** Neither the paid API release nor the AI Ultra release has one.
* **Other Gemini 4 models.** The September 30 announcement names only Argon. Searches for a "Gemini 4 Pro" or "Gemini 4 Flash" refer to models Google has not announced.
## Release questions
### Is Gemini 4 Argon out in the UK or Europe?
Google has not published a country-by-country schedule. Access is currently decided by the Fairwind Program, not by region.\[2]
### How do I use Gemini 4 Argon?
Today, only through the Fairwind Program if your organization is an approved cyber defense partner. Everyone else has to wait for the paid API and AI Ultra release.\[1]
### Is Gemini 4 Argon in the Gemini app or a Gemini subscription?
Not yet. Google lists Google AI Ultra subscribers among the first wider group to get access, but has not said when.\[1]
### Is Gemini 4 Argon the same as Gemini 4 Pro?
No. Gemini 4 Argon is the model Google announced. Google has not announced a model called Gemini 4 Pro.
### Is Gemini 4 Argon better than GPT-6 Astra?
On Google's own table it scores higher on most of the 18 benchmarks, and lower on FrontierSWE v2, Terminal-bench 4.0, Terminal-Bench Science and OSWorld-2.0.\[3] Independent results will need public access first.
### Can I use Gemini 4 Argon on SeedRouter?
Not yet. When it can be called through SeedRouter, the [Gemini 4 Argon page](https://seedrouter.ai/models/gemini-4-argon) will show its model ID, our live rate and the API reference. Until then, the same SeedRouter key works with the models already on the [models page](https://seedrouter.ai/models).
## References
1. Google. *Gemini 4 Argon: our next era of frontier intelligence.* September 30, 2026. Retrieved October 1, 2026 from [blog.google](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/).
2. Google DeepMind. *Fairwind Program.* Retrieved October 1, 2026 from [deepmind.google](https://deepmind.google/fairwind-program/).
3. Google DeepMind. *Gemini 4 Argon* (benchmark table). Retrieved October 1, 2026 from [deepmind.google](https://deepmind.google/models/gemini/).
4. Google AI for Developers. *Gemini models.* Checked October 1, 2026 at [ai.google.dev](https://ai.google.dev/gemini-api/docs/models).
5. Google DeepMind. *Gemini 4 Argon model evaluation: approach, methodology and results.* Retrieved October 1, 2026 from [deepmind.google](https://deepmind.google/models/evals-methodology/gemini-4-argon).