Gemini 4 Argon: How to Get Access (Plan, API Steps, Fairwind)

Gemini 4 Argon public release key facts: worldwide availability, API pricing and 1M-token limit

Gemini 4 Argon is public, but you need a paid plan to get it. Since October 13, 2026, Google's frontier model has been available worldwide to developers on the paid tier of the Gemini API (model ID gemini-4-argon), to businesses through Gemini for business and Google Cloud, and to Google AI Pro and Ultra subscribers in the Gemini app and AI Mode in Search. Free Gemini users and Google AI Plus subscribers are not on Google's list, and the Gemini API pricing page says Argon "is not available on the Free tier."

Updated October 20, 2026.

Access has widened once since launch. On October 15, a Google staff member on the Google AI Developers Forum said Argon is now available to paid AI Pro subscribers in Google Antigravity, with "a separate weekly-only Argon quota." Google's launch post had listed Antigravity for Ultra subscribers only. You no longer need the Fairwind Program to use the standard model. Fairwind now matters only to vetted cyber defenders who want Argon without cyber guardrails, which is a separate allowlisted model.

The price in one line: $2 per million input tokens and $10 per million output tokens through November 30, 2026, then $4 and $20. Our Gemini 4 Argon pricing guide covers the cached-input discount and what the December 1 change means for your bill.

How to get Gemini 4 Argon: step by step

Pick the route that matches how you want to use it. Steps 2 to 4 are for developers.

In the Gemini app or AI Mode in Search: subscribe to Google AI Pro or Ultra on the Google One plans page. Google says Argon comes with both plans "in select countries," and Gemini in Google AI plans is for users aged 18 and over. Free and AI Plus accounts are not on Google's list.
Get a Gemini API key: sign in to Google AI Studio. Google's Gemini API quickstart says new users get a project and key created automatically; otherwise click Create API key.
Switch that project to the paid tier: Argon is not on the free tier, so click Set up billing on the AI Studio API keys or Projects page, link a Cloud Billing account and, if asked, prepay at least $5 in credits, per Google's billing guide. The guide also says the $300 Google Cloud welcome credit can't be spent on the Gemini API for accounts opened after March 2, 2026.
Make your first call: install the SDK with pip install -U google-genai, set your key as GEMINI_API_KEY, then call client.interactions.create(model="gemini-4-argon", input="..."). The AI Studio developer guide sets thinking_level to "low", "medium" (the default) or "high".
In coding tools: in Google Antigravity, select Gemini 4 Argon as your active model (AI Ultra, and paid AI Pro since October 15 with a weekly Argon quota). In GitHub Copilot, GitHub says it is rolling out to Copilot Pro+, Max, Business and Enterprise users.
For the version without cyber guardrails: vetted defenders apply through Google DeepMind's Fairwind Program for gemini-4-argon-cyber-permissive, which is allowlist-only. Everyone else uses the standard model.

What happened

Google announced Argon on September 30 but gave it only to a set of Fairwind partners while it strengthened safeguards. On October 13, Koray Kavukcuoglu, SVP at Google DeepMind and Google's chief AI architect, wrote that Argon is "now globally available today" for developers, enterprises and paid Gemini subscribers. Google published separate posts for developers, the Gemini app, Google Cloud and Antigravity the same day, and GitHub made Argon generally available in Copilot.

The launch also changed what Fairwind is for. It is no longer the only way in. It is now the route to an unguarded version: Google says approved cybersecurity defenders can apply "to get access to Argon without cyber guardrails on a rolling basis," and it began expanding the program to hundreds of additional partners on launch day. The Fairwind Program page now says Google works with more than 650 partners globally.

Quick answers

What is the Gemini 4 Argon model ID?

The model ID is gemini-4-argon. The Gemini API docs list it as the stable version, and Google Cloud lists it as generally available from October 13, 2026. The cyber-defense version has its own ID, gemini-4-argon-cyber-permissive, and works only for allowlisted Fairwind partners.

Can free Gemini users use Gemini 4 Argon?

No. Google lists Argon for Google AI Pro and Ultra subscribers in the Gemini app and AI Mode in Search. Free users and AI Plus subscribers are not named. The Gemini API pricing page also says Argon is not available on the free tier, so developers need a paid API account to call it.

Which countries can use Gemini 4 Argon?

Google calls the rollout global but has not published a country list for Argon. Its Google One plans page says Argon is available to AI Pro and Ultra subscribers "in select countries." Google AI Pro and Ultra are each sold in more than 150 countries, and Gemini in Google AI plans is limited to users aged 18 and over.

Do I still need the Fairwind Program to use Argon?

Not for the standard model. Fairwind was the only way in from September 30 to October 13. It now serves vetted defenders, such as governments, critical infrastructure operators and core technology platforms, who want Argon without cyber guardrails. Google says it works with more than 650 Fairwind partners and is adding hundreds more.

What is the difference between Argon and Argon Cyber Permissive?

Cyber Permissive is a version of Argon for defensive security work such as vulnerability discovery, reverse engineering, exploit verification and patch generation. Google Cloud's docs say it is allowlist-only through Fairwind, defaults to HIGH thinking instead of MEDIUM, does not support computer use, and runs in the global and US regions but not the EU.

How do I get Argon in the Gemini app versus the API or Google Cloud?

For the Gemini app or AI Mode, subscribe to Google AI Pro or Ultra. Developers select gemini-4-argon on the paid Gemini API tier in Google AI Studio or Android Studio. Businesses use Gemini for business or Google Cloud's Gemini Enterprise Agent Platform. In Antigravity, Argon works for Ultra and, since October 15, paid AI Pro.

Do I have to prepay to use Gemini 4 Argon in the API?

Usually, yes. Argon needs the paid tier, and Google's billing guide says upgrading means linking a billing account and prepaying at least $5 in credits, with new users defaulting to Prepay. When the prepaid balance hits $0, every API key on that billing account stops working with an HTTP 402 error until you add credits.

How do I get more than 65,536 output tokens from Argon?

Use long decode. The AI Studio developer guide says the Interactions API defaults to 65,536 output tokens. To go further, set background=True, raise max_output_tokens above 65,536 and poll for the result. Input, thinking and output share one 1,048,576-token window, so the ceiling is about 1M tokens minus your input.

Who can use Gemini 4 Argon: access by plan

Where	Who gets Argon	Source
Gemini app and AI Mode in Search	Google AI Pro ("access") and AI Ultra ("higher access") subscribers, in select countries	Google; Google One
Free Gemini and Google AI Plus	Not listed by Google	Google; The Verge
Google Antigravity	AI Ultra from October 13; paid AI Pro from October 15, with a separate weekly-only Argon quota; eligible Gemini Enterprise subscriptions	Google; Antigravity blog; Google AI Developers Forum
Gemini API (Google AI Studio, Android Studio)	Paid tier only; not available on the free tier. AI Ultra subscribers can also try it in AI Studio	Gemini API pricing; Google developers post
Google Cloud (Gemini Enterprise Agent Platform)	Generally available; global, US and EU regions	Google Cloud docs
Gemini for business	Available on web, desktop and mobile	Google Cloud blog
GitHub Copilot	Copilot Pro+, Max, Business and Enterprise users (gradual rollout)	GitHub Changelog
Fairwind Program	Vetted defenders: Argon without cyber guardrails	Google DeepMind

Argon vs Argon Cyber Permissive

Google Cloud's Gemini 4 Argon documentation is the only official page that puts the two versions side by side. Here is what it says:

	Gemini 4 Argon	Gemini 4 Argon Cyber Permissive
Model ID	`gemini-4-argon`	`gemini-4-argon-cyber-permissive`
Access	Publicly available	Allowlist only (Fairwind Program)
Built for	Coding, finance, legal, cybersecurity and science work	Vulnerability discovery, reverse engineering, exploit verification, patch generation
Default thinking level	MEDIUM (LOW, MEDIUM, HIGH)	HIGH (LOW, MEDIUM, HIGH)
Computer use	Supported	Not supported
Regions	global, us, eu	global, us

The docs add that access to the earlier Gemini 3.8 Flash Cyber model does not carry over, so teams must apply again. The Fairwind page says partners may use Argon only for permitted dual-use work such as authorized threat simulation, reverse engineering and malware analysis. They must also use phishing-resistant MFA and limit access to internal security, incident response or penetration testing teams.

Developer details that trip people up

One token budget: Argon has a unified 1,048,576-token limit that covers input and output, including thinking and function calls. A single synchronous request returns up to 65,536 output tokens, and long decodes on the GenerateContent API can extend output toward the full limit.
Thinking levels: LOW, MEDIUM (the default) and HIGH. On Google Cloud, setting thinking_budget or thinking_level="MINIMAL" returns an HTTP 400 validation error. In the Gemini API, Google's developer guide says thinking_budget is deprecated and returns a warning, so switch to thinking_level either way.
Sampling and caching: custom temperature, top_p and top_k values are ignored, and only implicit caching is supported, with a 6,144-token minimum.
Lifecycle: Google Cloud lists a release date of October 13, 2026 and a retirement date of "October 13, 2027 or later" on its Agent Platform.
Tools: Google's AI Studio developer guide lists day-one support in the google-genai SDKs and in third-party tools including Vercel, LiteLLM, OpenRouter and Pydantic. GitHub Copilot bills Argon at provider list pricing under usage-based billing.

How good is it? Google's numbers vs independent tests

Google says Argon ranks #1 on the Vals Index. In its September 30 announcement it reported 77.9% on DeepSWE v1.1 and a tie for first at 68% on CWE-bench v1. The Verge notes that Google claims Argon beats GPT-6 Astra, Claude Fable 5.1 and Claude Opus 5.5 on benchmarks covering coding, cybersecurity, science, finance and law. Google ran those tests itself.

The independent picture is closer. Artificial Analysis scored Argon (high reasoning) 53 on its Intelligence Index, matching GPT-6 Astra (max), at $1.99 per index task, or 60% of Astra's cost at the discounted price. It also estimated that figure rises to $3.98, about 1.2 times GPT-6 Astra, once the discount ends. If you are weighing rivals, see our coverage of Claude Opus 5.5 Fast Mode and GPT-6's arrival in ChatGPT.

Inside Google, the Antigravity team says more than 100,000 people across Alphabet use Argon each month in an internal version of Antigravity. Over four weeks, it says, agentic code rose 35% and post-submit rollbacks fell 30%.

Long, multi-agent runs are where Google is pitching it hardest. In a paper posted to arXiv on October 15, Google DeepMind researchers report that four Argon agents beat Minecraft's Ender Dragon on their own in 27 hours and 14 minutes, writing 2,783 programs and dying 605 times along the way. The Next Web reports the researchers put the cost at about $520 at Argon's introductory price. The paper itself flags the limits: Easy difficulty, a bot library that tells agents what is around them, and a single run.

What Argon changes in the Gemini app

Google says Argon answers questions more directly "with less hedging" and shares its progress more often on long tasks. Alongside the model, the Gemini app gained a toggle from chat to tasks so Gemini can keep working while you are away, connections to more than 40 apps including Instacart, OpenTable, Adobe and Shopify, and Antigravity's coding interface in the Gemini macOS app. For work accounts, Google Cloud launched Argon a week after introducing its Gemini agent for work, and Argon is now available in Gemini for business on web, desktop and mobile.

Safety measures

Google says it trained Argon to respect system boundaries, report negative findings honestly and pause for human confirmation before irreversible steps. It says shortcut attempts on coding benchmarks fell to 2%. A separate automated monitor reads the agent's reasoning and actions to block unintended workarounds. Google also reports that external bio red-teamers could unlock less than 1% of the bio sub-tasks it tests for, and that Argon cut attack success on Gray Swan's indirect prompt injection benchmark to 0.7% across 15 adaptive attempts.

What's next

Google says "Gemini 4 Argon is just the beginning" and promises "the rest of the Gemini 4 family." There are three things to watch. The first is the December 1 price step, covered in our pricing guide. The second is whether Argon reaches AI Plus or free users, which The Verge says is unclear. The third is the size of the weekly Antigravity quota for AI Pro, which Google has not given a number for. The competition is moving fast too: Argon landed a day after Meta's Muse Wonder.

FAQ

Is Gemini 4 Argon available to the public?

Yes. Google made it globally available on October 13, 2026, for paid Gemini API developers, businesses, and Google AI Pro and Ultra subscribers in the Gemini app and AI Mode in Search. Before that it was limited to Fairwind partners.

How much does Gemini 4 Argon cost?

Through November 30, 2026, the Gemini API charges $2 per million input tokens and $10 per million output tokens, with cached input 95% off. From December 1 it costs $4 and $20. Consumers get it through a Google AI Pro or Ultra subscription.

Can I use Gemini 4 Argon in GitHub Copilot?

Yes. GitHub made it generally available in Copilot on October 13 for Copilot Pro+, Max, Business and Enterprise users, with a gradual rollout across VS Code, Visual Studio, JetBrains IDEs, Xcode, Eclipse, Copilot CLI and the coding agent. It is billed at provider list pricing.

What is the Fairwind Program?

It is Google DeepMind's limited-access cyber defense program for governments, critical infrastructure operators and core technology platforms. Partners get Argon without cyber guardrails for permitted defensive tasks, under rules such as phishing-resistant MFA and access limited to internal security teams.

How does Gemini 4 Argon compare with GPT-6 Astra?

Artificial Analysis scored Argon 53 on its Intelligence Index, matching GPT-6 Astra, at about 60% of Astra's cost per task under the introductory price. After the discount ends, it estimates Argon at about 1.2 times Astra's cost per task. Google's own benchmarks show Argon ahead.

How do I use Gemini 4 Argon in the API?

Create a key in Google AI Studio, set up billing on that project because Argon is not on the free tier, then call the model ID gemini-4-argon with the google-genai SDK or REST. Google's developer guide uses the Interactions API and a thinking_level of low, medium or high, with medium as the default.

Search This Blog

The Frontier Brief