Google’s most capable model yet, Gemini 4 Argon, is no longer limited to a handful of cybersecurity testers. Since 13 October 2026, access has opened to paid developers, Google AI Pro and Ultra subscribers, and enterprise customers — but the rollout comes with separate consumer and developer tracks, a limited-time price, and no free tier. Here is exactly who gets Argon, how each path works, and what it will cost you before and after 1 December.
Is there a Gemini 4 Argon waitlist?
No. As of 22 October 2026, Google has not announced any public waitlist for Argon. Access is eligibility-based rather than queued: if you already hold a qualifying product — a paid Gemini API key, a Google AI Pro or Ultra subscription, or an eligible enterprise plan — you can use it now. The only application-based route is the Fairwind Program, Google’s vetted scheme for cybersecurity defenders, which takes rolling applications for access without the model’s cyber guardrails. Everyone else either qualifies on the spot or does not have a path yet.
How Argon got here: a short timeline
- 2 September 2026 — Google opens the Fairwind Program alongside Gemini 3.8 Flash Cyber, its vetted-access scheme for cyber defenders.
- 30 September 2026 — Google DeepMind announces Gemini 4 Argon, its first Gemini 4-generation model, releasing it only to trusted cyber defenders through Fairwind.
- 13 October 2026 — General availability: the paid Gemini API, Google AI Studio, Antigravity, the Gemini app and AI Mode in Search all pick up Argon, with a public model ID (
gemini-4-argon) and an end date for the launch price.
Our launch-day coverage of the model itself is in Gemini 4 Argon: Google’s new frontier model explained; if you’re weighing it against the leaked Carbon checkpoint, read Argon vs Carbon: what is actually different.
The four ways to get Gemini 4 Argon access
Google’s rollout treats the consumer app and the developer API as separate products, and that distinction matters — AI Pro on the pricing page and AI Pro in the app are not the same thing, as pricing-watch site Folkfox has flagged. Pick your path:
1. The Gemini app and AI Mode in Search (consumers)
Google AI Pro ($19.99/month) and Ultra ($99.99/month) subscribers get Argon in the Gemini app and in AI Mode in Search. If you already pay for one of these plans, check your model picker — the rollout has been progressive, so it may appear without notice. Free-tier users and AI Plus subscribers do not get Argon at all; Google has said nothing about when, or whether, it reaches lower tiers. Desk-researched; not hands-on tested on a live Pro account.
2. The paid Gemini API and Google AI Studio (developers)
Every paying Gemini API customer can call Argon via model ID gemini-4-argon since 13 October 2026, and it is listed as stable in Google’s API documentation. It runs in AI Studio too, but note: in AI Studio, Pro subscribers need a paid API key — the subscription alone does not unlock the API. Argon is explicitly not on the API’s free tier, so the standard free-key experimentation route does not apply here.
3. Antigravity and the coding tools (builders)
Google Antigravity, the company’s agentic coding environment, offers Argon only to Ultra subscribers and to businesses through Gemini Enterprise. Related tooling follows the same line: Android Studio and Firebase AI Logic carry Argon for Android developers, and researchers can claim the headline offer of a top Antigravity usage tier free for six months — Google says it is extending that to 10,000 researchers.
4. Enterprise and the Fairwind Program (teams and defenders)
Enterprises go through Gemini Enterprise subscriptions (with contract, compliance terms and negotiated volume pricing, as multi-model gateway apimodels.app notes) or Vertex AI, where Argon has a Model Garden page — though Google has not named which regions host it, so EU data-residency planners should confirm before committing. Cybersecurity defenders can still apply to the Fairwind Program, which Google is expanding to hundreds of additional partners; this is the one route that strips the cyber guardrails.
Which access path fits you? Decision table
| You are… | Fastest path | What it costs | Watch out for |
|---|---|---|---|
| A curious individual | Google AI Pro in the Gemini app | $19.99/month | No guarantee it reaches Plus/free tiers; app and API are separate products |
| A developer prototyping | Paid Gemini API key + AI Studio | $2/$10 per million tokens until 30 Nov | Not on the free tier; thinking tokens bill at the output rate |
| An agentic-coding user | Antigravity via Ultra | $99.99/month | Pro does not qualify; Pro + API key only covers AI Studio |
| A company evaluating | Gemini Enterprise / paid API | Volume pricing, negotiated | Vertex AI hosting regions unconfirmed; budget at December rates |
| A cyber-defence researcher | Fairwind Program application | Rolling admission | Unguardrailed access; vetted-only |
Gemini 4 Argon pricing: what you pay, and when it changes
The launch price is a fixed window, not a sale: Google’s pricing footnote sets an end date of 30 November 2026.
| Tier | Input | Output | Cached input | From 1 December |
|---|---|---|---|---|
| Standard | $2.00 | $10.00 | $0.10 (95% off) | $4.00 / $20.00 / $0.20 |
| Batch & Flex | $1.00 | $5.00 | $0.05 | $2.00 / $10.00 / $0.10 |
| Priority | $3.60 | $18.00 | $0.18 | $7.20 / $36.00 / $0.36 |
Prices per million tokens, per Google’s API pricing page as reported by Tech AI Wire (Oct 2026). Three pricing realities that trip people up:
- The December price is the real price. At $4/$20, Argon matches Claude Opus 5.5 and Meta’s Wonder API list prices exactly; the intro rate matches GPT-6.1 Sol and Claude Sonnet 5.5 at $2/$10. Any project you start now will run mostly at December rates — re-price your estimate before you commit.
- Thinking tokens bill at the output rate. Argon always reasons: one report found a one-word “OK” reply consumed 66 thinking tokens. Set the thinking level to low for routine tasks and high only for the problems that need it, and read thinking-token counts from usage data rather than guessing from visible text.
- The 1M output limit has fine print. Google states a unified 1-million-token limit across input, thinking and output — but a normal request returns at most 65,536 output tokens; longer generations run as background jobs under the “continuation protocol” in the docs.
“Now globally available today” for developers and enterprises through the API, and for Ultra and Pro subscribers in the app — Koray Kavukcuoglu, Google’s October 13 announcement (via explainx.ai’s launch summary, Oct 2026)
While you wait: what to use instead
If you sit outside every access path today — free tier, Plus, or a region where the rollout has not landed — there is no waitlist to join, so waiting means re-checking, not registering. For the jobs Argon is pitched at, these are the honest stand-ins, each with a pricing anchor you can verify on the vendors’ pages:
- Long-context reasoning on a budget: Gemini 3.8 Flash is priced at $0.75/$3.75 per million tokens through 31 December 2026 — the cheap way to test your 1M-context pipelines before Argon’s December price bites.
- Agentic coding parity: At intro prices, Argon matches GPT-6.1 Sol and Claude Sonnet 5.5 token-for-token ($2/$10) — but independent tests put Argon level with GPT-6 Astra at 53 on the Artificial Analysis Intelligence Index and behind Claude Opus 5.5 (58), so the cheaper rival is not automatically worse for your task.
- Third-party gateways: Multi-model gateways such as apimodels.app listed Argon at $1.10/$5.50 until 30 November (55% of Google’s list, checked 17 Oct 2026) — a cheaper route if you can accept another company in your data path and contract terms that differ from Google’s.
Bottom line: check your eligibility first — many people hunting for a waitlist already qualify through a Pro subscription or a paid API key. Budget every estimate at $4/$20, test inside the cheap window, and treat anything billed at $2/$10 as a one-off saving, not the new normal.
Frequently Asked Questions
Can I use Gemini 4 Argon for free?
No. Google lists Argon as unavailable on the API’s free tier and on the free Gemini app tier. The cheapest route in is a Google AI Pro subscription at $19.99/month for app access, or a paid API key at the intro token rates.
Is Gemini 4 Argon available worldwide?
Google describes the 13 October rollout as global, but its own posts name no regions — Reuters characterises the release as worldwide. EU data-residency users should confirm Vertex AI hosting regions before relying on it for regulated workloads.
Do I need to join a waitlist for Gemini 4 Argon?
No public waitlist exists. You either hold a qualifying product (paid API key, AI Pro/Ultra, enterprise plan) or you apply to the Fairwind Program as a cyber defender. There is nothing to register for as a consumer.
What happens to pricing on 1 December 2026?
The introductory API price expires on 30 November 2026. From 1 December, standard rates double to $4 per million input and $20 per million output tokens, with cached input rising from $0.10 to $0.20.
How does Argon compare to Gemini 4 Carbon?
Carbon is a leaked internal checkpoint on the Jetski training run (Business Insider, 9 Oct 2026) — not a shipping product — while Argon is generally available. See our full breakdown in Gemini 4 Argon vs Carbon: what is actually different.
Sources and methodology
Access and pricing claims verified against secondary reporting of Google’s 13 October 2026 announcement on 22 October 2026: Tech AI Wire’s GA pricing breakdown (dev.to), explainx.ai’s launch summary (explainx.ai), Moneycontrol’s launch report (Moneycontrol), datanorth.ai’s access guide (datanorth.ai), plus the airankings model page and the apimodels.app pricing note (via Medium, checked 17 Oct 2026). Google’s own pricing pages and developer docs were not directly re-checked this run; prices are labelled per Google’s published footnote as reported by the sources above. Desk-researched; not hands-on tested. Verification date: 22 October 2026.
Govind Dheda covers AI tools, guides and prompts for OpenAIMaster — what is new, what is worth using, and how to put AI to work.