How to use Gemini 4 Argon: in the Gemini app, AI Studio, the Gemini API and Antigravity

How to use Gemini 4 Argon depends on which door you go through. Google made its newest Gemini model generally available on October 13, 2026 (our news story covers the launch), but each surface has its own requirement. The Gemini app needs a paid Google AI plan, AI Studio and Antigravity need the top plan, and the API has no free tier at all. This guide covers what you need for each, the setup steps, what it costs and what to do when Argon doesn't show up.

The steps below come from Google's documentation as of October 14, 2026, mainly the "What's new in Gemini 4 Argon" developer page (last updated October 13, 2026 UTC), the Gemini API pricing, rate-limit and terms pages, and Google's help and plan pages. Code adapted from Google's Gemini API documentation as of October 14, 2026; it was not run for this guide. The code blocks were compared line by line with the live "What's new" page on October 14, 2026 and match it (the thinking-level snippet is an excerpt of Google's longer sample).

The short version

• Gemini app and AI Mode in Search: Google AI Pro or Ultra.
• Google AI Studio: Ultra for subscription access in the Playground, or any account using a billing-enabled API key.
• Gemini API: a paid tier (a project with Cloud Billing). Model ID: gemini-4-argon.
• Antigravity: Ultra for individuals; businesses through a Google Cloud plan.
• Price: $2 / $10 per million input/output tokens through November 30, 2026; $4 / $20 from December 1.

Which plan do you need for Gemini 4 Argon?

Last verified: October 14, 2026 (gemini.google subscriptions page, ai.google.dev Google AI plans and pricing pages, antigravity.google pricing page)

Google's launch post, its Gemini app release notes and the banner on its plan page ("Select AI Pro or AI Ultra below to get access") name only AI Pro and Ultra. Free accounts don't get Argon. AI Plus ($4.99) is unclear: on October 14, Google's plan pages showed an "Includes Gemini 4 Argon" line at the end of the AI Plus card, but the Plus feature list itself names only Google's Flash model. Treat Plus access as unconfirmed. Google says its AI plans are sold in more than 140 countries and territories, and that some AI benefits, including AI Studio, are for users 18 and over.

How do you use Gemini 4 Argon in the Gemini app?

1. Subscribe to AI Pro or Ultra through Google One, or from the upgrade option in the Gemini app.
2. Open Gemini at gemini.google.com, or in the mobile or desktop app, signed in with the subscribed account.
3. Pick the model. Google's help center says that on the web you click the model name inside the text box at the bottom of the page and choose a model. On mobile, 9to5Google reported in October 2025, the picker sits in the prompt bar. Select Gemini 4 Argon.

Google's release notes say Argon is "rolling out globally" from October 13, so it may take time to reach every account. Google hasn't published numeric Argon limits for the app; Ultra gets "higher access" than Pro. In AI Mode in Search, Google lists Argon access for Pro and Ultra but hasn't documented how to choose it.

How do you use it in Google AI Studio?

AI Studio has two routes:

• With an Ultra subscription: Google's AI plans page says Pro and Ultra subscriptions "unlock paid models and higher rate limits" in the AI Studio Playground, with daily quotas, and that Argon "is not available on the Free or Google AI Pro plans." Upgrade from the button in AI Studio's left navigation, or on one.google.com. Then choose Gemini 4 Argon in the Playground's model list.
• With a paid API key: once your daily subscription quota runs out, or if you don't have Ultra, you can keep working with a Gemini API key on a project with Cloud Billing, paying per request.

Subscription benefits apply only inside the AI Studio web interface. Direct API calls are billed separately.

How do you call Gemini 4 Argon from the Gemini API?

Prerequisites: a Google Cloud project with an active billing account (Tier 1 or higher), a Gemini API key, and Google's Gen AI SDK. Per the quickstart, install it with pip install -U google-genai (Python) or npm install @google/genai (JavaScript), and set the key as an environment variable with export GEMINI_API_KEY="YOUR_API_KEY".

Google's "What's new" page uses the Interactions API. Its Python quickstart for Argon:

from google import genai

client = genai.Client()

interaction = client.interactions.create(
    model="gemini-4-argon",
    input="Write a three.js script that renders a realistic 3D black hole with gravitational lensing."
)

print(interaction.output_text)

The same request over REST, from the same page:

curl "https://generativelanguage.googleapis.com/v1beta/interactions" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H 'Content-Type: application/json' \
  -X POST \
  -d '{
    "model": "gemini-4-argon",
    "input": "Write a three.js script that renders a realistic 3D black hole with gravitational lensing."
  }'

To set reasoning effort, Google's thinking-level sample adds this argument inside the create call:

    generation_config={
        "thinking_level": "medium"  # Balanced reasoning effort for complex tasks
    }

The levels are low, medium (the default) and high. Google recommends evaluating all three and using the lowest that meets your quality bar. Our reasoning models explainer covers why: thinking tokens are billed as output.

What changes if you're migrating from an older Gemini model, per Google's checklist:

• temperature, top_p and top_k are ignored. Remove them.
• Explicit context caching isn't supported. Remove cachedContents, caches.create and cached_content. Implicit caching is on by default for prompts of at least 4,096 tokens.
• If you set max_output_tokens explicitly, raise it to at least 32,768, because it covers thinking and the answer together and is a hard limit. The default is 65,536.
• Raise HTTP client timeouts to at least 15 minutes.
• Use previous_interaction_id to keep multi-turn and function-calling state on the server.

Long outputs. Argon's 1,048,576-token window is shared by input, thinking and output. A normal request returns up to 65,536 tokens. For more, Google's docs say to set background: true and a max_output_tokens above 65,536. Its long-decode sample uses "max_output_tokens": 262144. The model then writes in hops of up to 64K tokens, and you have to join the text from every model_output step.

Argon accepts text, image, video, audio and PDF input and returns text only. Google's model page lists function calling, code execution, search grounding, structured outputs, URL context and file search. It doesn't support image generation, audio generation or the Live API.

How do you use Gemini 4 Argon in Antigravity?

Antigravity is Google's agent-first coding app, available for macOS, Windows, Linux and Googlebook. Google's launch post gives individual Argon access in Antigravity to Ultra subscribers. Antigravity's own October 13 post says Ultra subscribers get it in the desktop app and the CLI, and that it is also available on eligible Gemini Enterprise subscriptions; businesses can also use it through Google Cloud plans.

1. Download Antigravity from antigravity.google, or use the Antigravity button in the Gemini macOS app. Google's October 13 release notes say that option is for personal accounts and eligible work accounts aged 18+, in English.
2. Sign in with the Google account that holds the Ultra subscription.
3. Choose Gemini 4 Argon as the agent model. Antigravity's site carries a "Gemini 4 Argon available in Google Antigravity" banner, but Google's launch posts, Antigravity's October 13 post and its pricing page don't describe how to pick it.

Developers using Google's Managed Agents through the API can point the Antigravity agent at Argon by setting "model": "gemini-4-argon" in agent_config, per the "What's new" page. Scientists can also apply for a Google offer of the top Antigravity usage tier free for six months, open to 10,000 people subject to eligibility and capacity.

What does Gemini 4 Argon cost per request?

This is an illustrative calculation from list prices, not a measured bill. Output includes thinking tokens, which vary by thinking level.

Last verified: October 14, 2026 (ai.google.dev pricing page)

At 1,000 uncached calls like the first row a day, that's about $300 a day now and $600 from December. Budget on the standard price. Our explainer on why token prices mislead covers the other factors.

What are the limits and quotas?

• Rate limits: Google's rate-limit page doesn't publish per-model Argon numbers. It says to check them in AI Studio, where they depend on your tier.
• Tiers: Tier 1 needs a linked billing account. Tier 2 needs $100 paid and 3 days since the first payment. Tier 3 needs $1,000 paid and 30 days.
• Spend-based limits: some accounts also face a spend cap per rolling 10-minute window ($10 at Tier 1, $50 at Tier 2, $200 at Tier 3) and get a 429 RESOURCE_EXHAUSTED error past it. At the introductory price, a single 1M-token output costs $10.
• Batch: Argon's enqueued-token limit is 5 million at Tier 1, 500 million at Tier 2 and 1 billion at Tier 3.

How is your data used?

• Gemini API (paid): Google's terms say it doesn't use your prompts or responses to improve its products. It logs them "for a limited period of time" to detect policy violations.
• AI Studio: the terms treat direct AI Studio use as unpaid, which means Google may use content to improve products and human reviewers may read it, unless the account is tied to a billing-enabled Cloud project. The terms don't say how an Ultra subscription changes this, so don't paste confidential data into AI Studio on a personal account.
• Gemini app: with Keep Activity on, Google says it uses chats to improve its services, "including training generative AI models," and a subset is read by human reviewers. Activity is deleted after 18 months by default. With it off, new chats are kept for 72 hours.

Troubleshooting: why can't I see Gemini 4 Argon?

• Not in the Gemini app: check that you're on AI Pro or Ultra and signed in with that account. The rollout is gradual, and plan availability varies by country. Free accounts don't get Argon, and Google hasn't confirmed AI Plus access.
• Not in the AI Studio Playground: Pro doesn't include Argon there. Use Ultra or a billing-enabled API key. AI Studio is 18+.
• API errors with a free key: Argon isn't on the free tier. Link a billing account to the key's project.
• Caching errors after migrating: remove explicit caching calls.
• Truncated or cut-off answers: raise max_output_tokens, or use background execution for anything over 64K tokens.
• Timeouts: raise client timeouts to 15 minutes or more, as Google recommends.
• Refusals on security work: the public model blocks offensive tasks such as pentesting and exploitation, per Google Cloud. Vetted defenders can apply to the Fairwind Program, which is how Argon first shipped (our September 30 story).

What's confirmed, reported and uncertain

• Confirmed (Google docs, October 13–14): plan requirements, model ID, prices and the November 30 cutoff, no free API tier, the unified 1,048,576-token window, migration rules, tier qualifications and data-use terms.
• Uncertain: whether Google AI Plus includes Argon, numeric Gemini app limits, per-model API rate limits (shown only in AI Studio), how Argon is chosen in AI Mode and Antigravity, how an Ultra subscription affects AI Studio data use, and when free users get Argon.