Gemini 4 Argon is Google’s frontier model, announced on September 30, 2026 and generally available since October 13, 2026. Developers call it as gemini-4-argon on the paid tier of the Gemini API, and Google AI Pro and Ultra subscribers get it in the Gemini app and Search. It costs $2 per million input tokens and $10 per million output tokens through November 30, 2026, then $4 and $20 from December 1. There’s no free API tier.
This guide is the hub for our Argon coverage: specs, access by channel, cost, benchmarks, and a first request. Choosing a model? Read our Gemini 4 Argon vs GPT-6 Astra vs Claude Opus 5.5 comparison. If you test model APIs in Apidog, the last section shows how to set up an Argon request with a cost check attached.
What Google has confirmed vs what it hasn’t
Most of the gaps from launch week are now filled. Here’s the split as of October 15, 2026, based on the Gemini API changelog, the Argon model page and Google’s pricing page.
| Item | Status | What we know |
|---|---|---|
| Announcement | Confirmed | September 30, 2026, Fairwind Program partners first |
| General availability | Confirmed | October 13, 2026 (changelog: “generally available”) |
| Model ID | Confirmed | gemini-4-argon, stable, no -preview suffix |
| Price | Confirmed | $2/$10 per 1M tokens through November 30, then $4/$20 |
| Context window | Confirmed | 1,048,576 tokens, shared by input, thinking and output |
| Output limit | Confirmed | 65,536 by default; up to about 1M with long decode |
| Thinking levels | Confirmed | low, medium (default), high |
| Free API tier | Confirmed: none | Paid tier only |
| Per-model rate limits | Not in the docs | Shown in AI Studio for your project |
| Knowledge cutoff | Not published | Nothing from Google |
| Vertex AI / Agent Platform | Not confirmed | Don’t assume it’s there yet |
Argon takes text, image, video, audio and PDF input and returns text only. It supports function calling, structured outputs, code execution, grounding, computer use (Preview) and the Batch API, but not explicit caching, the Live API, or audio and image generation.
What Gemini 4 Argon is and where it sits
Argon is the only Gemini 4 model announced so far and the new top of Google’s lineup. Google calls it “our new frontier model,” built “to sustain deep reasoning across complex, long-horizon workflows,” and claims frontier performance in software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense. A Google spokesperson told Reuters that Argon is larger than the company’s previous line of “Pro” models.
Before Argon, the top Gemini API model was gemini-3.1-pro-preview, still in Preview, at $2 input / $12 output per million for prompts up to 200K and $4/$18 above that. Argon’s $10 output rate through November 30 undercuts it, and Argon has no long-prompt tier. Reuters said the launch came “after months of delays,” after a promised June release of Gemini 3.5 Pro was dropped.
Google also runs Argon internally: it says thousands of employees use it, and that Argon agents freed over 300 TiB of fleet memory.
Who can use Gemini 4 Argon, and where
On October 13, Sundar Pichai said Argon is “available globally today for developers and enterprises in the paid tier of the Gemini API and the Gemini platform, for Ultra subscribers in Google Antigravity, and Ultra and Pro subscribers in the Gemini app and Search.”
| Channel | Who gets Argon (as of October 15, 2026) |
|---|---|
| Gemini API | Paid tier only, model gemini-4-argon |
| Gemini app | Google AI Pro and Ultra subscribers |
| Search | Google AI Pro and Ultra subscribers |
| Google Antigravity | Ultra subscribers |
| Cursor | Added October 13, at $2/$10 per 1M tokens |
| OpenCode Go | Added October 14 |
| Fairwind Program | A set of partners since September 30 |
There’s no free route on the API. Google’s pricing page says “Gemini 4 Argon is not available on the Free tier or on the Google AI Pro plan.” That line is about API plans; Pro subscribers still get Argon in the Gemini app. A consumer subscription isn’t an API key, and Google hasn’t published per-plan message limits for the app. Our release date guide has the full timeline, and is Gemini 4 Argon free covers cheaper options.
The rollout started with Google’s limited-access Fairwind Program for cyber defenders, who use Argon as a managed model on Gemini Enterprise, on its own or inside CodeMender.
Gemini 4 Argon pricing
| Per 1M tokens | Through Nov 30, 2026 | From Dec 1, 2026 |
|---|---|---|
| Input | $2.00 | $4.00 |
| Output, including thinking | $10.00 | $20.00 |
| Cached input (implicit) | $0.10 | $0.20 |
| Batch / Flex input and output | $1.00 / $5.00 | $2.00 / $10.00 |
| Priority input and output | $3.60 / $18.00 | $7.20 / $36.00 |
There’s no cache storage fee, because explicit caching isn’t supported. Implicit caching is on by default for prompts of 4,096+ tokens and takes 95% off cache hits. From December 1, Argon’s list price matches Claude Opus 5.5; through November 30 it matches GPT-6.1 Sol and Claude Sonnet 5.5 (vendor pages, checked October 1). Worked scenarios are in Gemini 4 Argon pricing.
The 1M context window and long decode
At launch, Google said it was expanding “the model’s output token limit to an industry-leading 1M tokens, up from the previous 64K tokens.” The docs now spell out how that works. Argon has a unified 1,048,576-token context window that input, thinking and output share. A synchronous request returns up to 65,536 output tokens by default. To go further, you use long decode: an Interactions API call with background: true and max_output_tokens above 65,536, up to 1,048,576 minus your input tokens.
The model generates in hops of up to 64K, and the server runs the continuation hops for you; you join the text from every model_output step. Hard prompts take minutes, so use background mode or set HTTP timeouts to at least 15 minutes. Our guide to Argon’s 1M output tokens covers timeouts and cost caps.
Gemini 4 Argon benchmarks: the headline
All numbers in this table come from Google’s launch post and evals methodology. Argon’s scores are Google-reported, some self-computed and some from leaderboards such as Vals AI; competitor scores are mostly the vendors’ own or public-leaderboard results, and harnesses differ. By our recount, Argon leads 13 of 19 rows outright, ties one and trails on five.
| Benchmark | Gemini 4 Argon | GPT-6 Astra | Claude Fable 5.1 | Claude Opus 5.5 |
|---|---|---|---|---|
| Vals Index | 68.9% | 63.1% | 65.8% | 67.0% |
| DeepSWE v1.1 | 77.9% | 74.1% | 67.4% | 74.2% |
| GraphWalks 256K to 1M (F1) | 84.2% | 71.8% | 65.0% | 66.8% |
| FrontierSWE v2 | 55.0% | 65.5% | 56.3% | 62.3% |
| Terminal-bench 4.0 | 57.4% | 58.2% | 57.9% | 66.4% |
| PostTrainBench | 45.3% | 44.3% | 40.2% | 49.3% |
Since GA, outside numbers have landed. Cursor reports 52.8% on CursorBench at default effort, “on par with Opus 5.5.” Scale AI puts Argon first on Humanity’s Sixth Sense at 55.4%, ahead of GPT-6 Astra at 53.6%. As of October 15, Artificial Analysis scores Argon (High) 53 on its Intelligence Index, #8 of 216 variants; in launch-week data that tied GPT-6 Astra and Claude Fable 5.1 at max settings, behind Claude Opus 5.5 (58) and Claude Sonnet 5.5 (56). All 19 rows and Google’s method notes are in Gemini 4 Argon benchmarks.
Cyber defense in one paragraph
Google says “Argon can autonomously find, validate, and patch critical software vulnerabilities,” and at announcement it said trusted defenders would get a build without cyber guardrails. On DeepMind’s cyber page, Argon scores 85.8% on real-world vulnerability discovery (Gemini 3.8 Flash Cyber: 71.0%), ties Grok 4.7 and GPT-6 Astra at 68% on CWE-bench v1, and posts the chart’s lowest Gray Swan indirect prompt injection attack success rate, 0.7% at k=15. See what Argon’s cyber release means for the APIs you run and its predecessor, Gemini 3.8 Flash Cyber.
How to call Gemini 4 Argon
Google’s developer guide uses the Interactions API for Argon. Keep the model name in one variable, so you can compare against gemini-3.8-flash or roll back without a code change.
MODEL="${GEMINI_MODEL:-gemini-4-argon}"
curl -s "https://generativelanguage.googleapis.com/v1beta/interactions" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H "Content-Type: application/json" \
-X POST \
-d '{
"model": "'"$MODEL"'",
"input": "List the breaking changes in this OpenAPI diff.",
"generation_config": {"thinking_level": "medium"}
}'
Moving Gemini 3.x code over? Google’s migration checklist says temperature, top_p and top_k are ignored, explicit caching (cachedContents, caches.create) is gone, and multi-turn flows should use previous_interaction_id.
In Apidog, save the call as a request with GEMINI_API_KEY and GEMINI_MODEL as environment variables, then assert on the token counts in the response’s usage object (total_output_tokens plus total_thought_tokens, since thinking bills as output) so a runaway answer fails the test. Save a gemini-3.8-flash response as your baseline, then switch GEMINI_MODEL and compare quality, tokens and cost.
For request shapes in Python and long decode, see the Gemini 4 Argon API guide. To decide whether Argon is worth the price over Flash, read Argon vs Gemini 3.8 Flash.
FAQ
Can I use Gemini 4 Argon in the Gemini API? Yes, on the paid tier since October 13, 2026. There’s no free tier for Argon.
What is the Gemini 4 Argon model ID? gemini-4-argon. It’s the stable version and the only one listed; there’s no -preview suffix.
When was Gemini 4 Argon released? Google announced it on September 30, 2026, for Fairwind partners, and made it generally available on October 13, 2026.
Is Gemini 4 Argon better than GPT-6 Astra and Claude Opus 5.5? It leads most rows in Google’s table and tops Scale AI’s Humanity’s Sixth Sense, but Artificial Analysis’s launch-week data had it level with Astra and behind Opus 5.5. For the rivals on their own terms, see what is Claude Opus 5.5 and our GPT-6 Astra API guide.
Does Gemini 4 Argon have a 1M context window? Yes: 1,048,576 tokens, shared by input, thinking and output. The first version of this guide treated 1M as an output limit only; Google’s model page now documents it as a unified context window.
Can Google AI Pro subscribers use Argon? In the Gemini app and Search, yes. Not through the API: a Pro plan isn’t an API key, and the pricing page excludes it.
Start building on Argon
Argon is live on the paid Gemini API, so the work now is control, not waiting. Put the model in a variable, pick the thinking level on purpose, and attach a token assertion to every request. Price those tests at the December 1 rates ($4/$20) so your budget holds when the rate changes. Need a cheaper fallback? Our Gemini 3.8 Flash API guide covers it. To keep the request, environment and baseline in one place, download Apidog.



