Who Can Use Gemini 4 Argon Now? Fairwind, API, and Ultra Access
As of October 2, 2026, Gemini 4 Argon is not generally available. Google confirms a limited rollout to selected trusted cyber defenders through Fairwind. Paid API customers and Google AI Ultra subscribers are named as the starting groups for a later broader release, but no public date, model ID, or universal entitlement has been announced. This guide explains the access paths, verification steps, pricing conditions, and when to keep using an available model.
Contents
You may have seen the Gemini 4 Argon announcement and then found no Argon option in AI Studio, the API model list, or the Gemini app. The practical question is not how impressive the launch sounds; it is whether your account can use the model today and whether a real project should wait for it.
Bottom line, last checked October 2, 2026: Google only confirms that Argon is rolling out to a set of trusted cyber defenders through the Fairwind Program. Google says the later, broader release will start with paid API customers and Google AI Ultra subscribers, but it has not announced a public date or an order between those two groups. Paying for API usage or subscribing to Ultra does not, by itself, prove that your account has Argon.
Current access matrix: who can use Argon and who still has to wait
| Access path | Status on October 2, 2026 | What that status actually means | What to do now |
|---|---|---|---|
| Selected trusted cyber defenders in Fairwind | Confirmed limited rollout | Argon is being assigned to a set of defenders, not automatically to every applicant or partner | Treat it as available only after an explicit grant and access in the managed environment |
| New Fairwind applicants | Program applications are open; Argon is not guaranteed | Applying, being admitted, and receiving a specific frontier model are separate events | Apply only if your organization fits the program and wait for explicit model authorization |
| Paid Gemini API customers | Named as a starting group for the next broader release; not confirmed as generally live | “Paid” describes rollout priority, not a current account entitlement | Check the model list for your own project; keep using an available model until Google publishes an exact ID and your account returns it |
| Google AI Ultra subscribers | Named as a starting group for the next broader release; not confirmed as generally live | An Ultra receipt is not evidence that Argon is in your Gemini model selector | Verify the model inside the correct signed-in account and start a real session |
| Other developers, enterprises, and consumers | Planned later; date unknown | Google says access will expand gradually but gives no complete schedule, regions, or product surfaces | Do not commit a deadline-dependent project to Argon yet |
The central distinction is simple: a group named in a future rollout is not the same as a group that can use the model now. Google’s official Argon announcement uses phased-release and future-access language.
Who Fairwind early access is designed for
Fairwind is not a normal consumer waitlist. It is a controlled program intended to give high-priority cyber defenders early access to frontier capabilities. According to the official Fairwind page, Google currently prioritizes:
- governments and national cyber authorities;
- critical-infrastructure operators, including healthcare, telecommunications, energy, and financial networks;
- core technology platforms that protect software foundations used by many downstream users.
Academic labs focused on defensive benchmarking may also apply. Google says it vets applicants for security history and ethical operations. Participating organizations must accept governance controls such as user-level authentication, phishing-resistant multifactor authentication, access tracking, and restrictions against sharing, redistributing, or selling access. Permitted dual-use work includes authorized threat simulation, reverse engineering, and malware analysis for defensive or academic research; malicious uses are prohibited.
A Fairwind application is not an instant Argon grant
The Fairwind page has an “Apply for access” route, but that form requests admission to the program; it is not a public one-click Argon waitlist. The current page now names Argon directly: it says a set of Fairwind partners get exclusive access, the FAQ says the program provides exclusive Argon access, and the governance terms specify which internal teams may receive it. The limits still matter—“a set” is not every partner, and the program is for approved trusted partners—so use three separate checkpoints:
- your organization is approved for Fairwind;
- Google explicitly tells your organization that Argon is included;
- Argon appears in the assigned managed environment and works under the program terms.
Until all three are true, do not translate “we applied to Fairwind” into “we have Argon.” Individual developers and students should not assume that a personal research idea qualifies; the public criteria prioritize organization-level defensive missions.
Will paid API customers or Ultra subscribers get access first?
Google has not announced an order between them. The release says broader access will start with “paid API customers and Google AI Ultra subscribers.” That wording puts both groups in the same announced starting wave; it does not say API first, Ultra second, or the reverse.
Three common assumptions are therefore unsafe:
- “My Gemini API project has billing enabled, so Argon must be callable now.”
- “I bought Ultra, so Argon must already be in the Gemini app.”
- “Google published a launch price, so an API endpoint must already be active.”
The announcement does not provide a public model ID, a general Argon signup link, a region list, or a product-by-product sequence for AI Studio, Vertex AI, the Gemini API, and the Gemini app. Wait for an official product surface or an account entitlement before treating a path as live.
How to verify API access without guessing a model ID
Do not use a forum screenshot, plan name, or headline as proof. For a Gemini API project, the strongest practical evidence is that the complete model list returns an officially documented Argon model ID for your project and that the ID completes a minimal request. The official Models API reference says models.list lists models available through the Gemini API, but the REST method returns only 50 models per page by default and may include a nextPageToken. A single /v1beta/models request therefore checks only the first page; “Argon was not in that response” is not an entitlement result.
Install Google’s official google-genai SDK, read the API key without displaying it, and iterate the model list through the SDK:
python3 -m pip install --upgrade google-genai
read -rs GEMINI_API_KEY && export GEMINI_API_KEY
python3 - <<'PYCODE'
from google import genai
client = genai.Client()
for model in client.models.list():
print(model.name)
PYCODE
unset GEMINI_API_KEYThe official SDK iterator handles the paginated result. If you use REST instead, pass each response’s nextPageToken as the next request’s pageToken until no token is returned.
After the complete iteration finishes, verify in this order:
- Look for a name or ID that Google has formally documented as Argon in the complete result; do not invent
gemini-4-argon. - Confirm that the returned entry supports the generation method you need and exposes the relevant limits.
- Use the exact ID returned by Google for a minimal request and make sure the result is not a
404, permission error, or regional-access error. - Only then review quota, billing, data terms, and production readiness.
Keep public-catalog status separate from account results. When this article was checked, the public Gemini API model catalog did not list Argon; that only means the public guide had not added it, not that your account was necessarily unauthorized. Conversely, if a complete models.list iteration does not return Argon, it means the current credential has no verifiable Argon entry—not that you should guess an ID. The final proof is still an official ID returned by Google plus a successful minimal call.
Ultra subscribers need an entitlement check too
For Ultra, a subscription receipt is not the test. Sign in to the exact Google account that owns the plan, open the Gemini model selector, confirm that Argon is explicitly listed, and start a new session. The current Google AI plan page describes Ultra’s existing benefits; the Argon announcement only names Ultra subscribers as one starting group for a future broader release.
If Argon is absent, record the status as “not enabled for this account yet.” Reinstalling the app, switching interface language, or buying a duplicate plan does not create an entitlement that Google has not granted.
When the announced Argon price actually matters
Google announced a future introductory launch price of $2 per million input tokens and $10 per million output tokens, with cached input priced at a 95% discount to the input-token price. A footnote says the post-introductory price will be $4 per million input tokens and $20 per million output tokens.
Those numbers do not prove that the public API is live. Google has not published the date on which the introductory period starts, how long it lasts, or whether every product surface begins billing at the same time. Use the figures as announced future rates for planning, not as a currently purchasable service, until your account shows the model, an applicable price page, and working billing rules.
Should you wait for Argon or keep using an available model?
You have a delivery deadline: use a model that is available now
For a launch due in days or weeks, choose a model that appears in the current official catalog, accepts requests, has known quota, and can be billed in your project. Put the model ID in configuration rather than hard-coding it, preserve an adapter boundary, and prepare a representative evaluation set. When Argon arrives, you can run an A/B test instead of rebuilding the system.
Do not promise customers Argon’s quality, latency, regional coverage, or quota based on launch benchmarks and internal Google examples. None of those production conditions can be verified in your environment before access exists.
Your organization fits Fairwind: apply now, but keep the current defense stack
Governments, critical-infrastructure operators, core technology platforms, and defensive-evaluation labs can submit the Fairwind form. Keep the existing security workflow running while the application is reviewed: Google publishes no processing deadline, and program admission does not guarantee immediate access to every frontier model.
A useful application should state the defensive use case, protected systems, authorization boundary, data-governance plan, approved users, and security controls. That aligns better with Fairwind’s published criteria than a generic request to try the newest model.
You already pay for API or Ultra and have no urgent deadline: prepare an evaluation, not a purchase gamble
Build a set of 20–50 real tasks, success criteria, a cost-recording method, and a safety-review process. Once Argon appears in your account, compare it with the model you use today on the same tasks. Until entitlement is visible, do not upgrade a plan or prepay more API balance solely in the hope of earlier access; Google has not promised that a purchase action unlocks Argon immediately.
Your task truly depends on an Argon-specific capability: label the external dependency and design a fallback
If the research design genuinely depends on a capability Google has announced for Argon, mark the project as waiting on an external dependency. In parallel, define a fallback: narrow the task, split it into stages, use an available model with human review, or postpone validation until official access arrives. Avoid unverified third-party gateways for sensitive code or security data.
Frequently asked questions
Is Gemini 4 Argon publicly released?
The model has been announced, but general public availability has not been confirmed. The currently confirmed access is a limited rollout to selected trusted cyber defenders in Fairwind.
Does Google AI Ultra include Argon today?
Do not infer that from the subscription alone. Ultra subscribers are named as one starting group for the next broader release; your account’s model selector must actually show Argon.
Do paid Gemini API projects have Argon now?
Google has not announced access for every paid project. Iterate the complete paginated models.list, then make a successful minimal call with an official ID returned by Google; do not rely on the first page or a guessed ID.
Is there a public Fairwind application?
Yes. Eligible defensive organizations can apply on the Fairwind page, but application review, program admission, and Argon authorization are separate steps with no published processing time.
Which comes first, API or Ultra?
There is no verifiable order yet. Google names both in the same “starting with” group.
Has Argon billing already started?
The published figures are launch and post-introductory rates. The public launch date and the length of the introductory period are not stated.
Final decision
As of October 2, 2026, the accurate planning assumption is controlled early access, not general availability. Eligible Fairwind organizations can apply; paid API and Ultra users can prepare evaluations and monitor their account entitlements; teams with deadlines should keep shipping on an accessible model and make a later Argon switch configurable and reversible rather than basing the plan on an undated release.
Status checked October 2, 2026. Access can change; confirm against Google’s Argon announcement, the Fairwind page, the Gemini API model catalog, and the entitlement shown in your own account.
