No. Gemini 4 Argon, the model Google released on October 13, 2026, isn’t in the free Gemini app, the free API tier, Google AI Plus or the student plan. You get it with Google AI Pro or Ultra, or by paying per token on the API: $2 in and $10 out per 1M tokens, doubling on December 1. Even on Pro it isn’t unlimited. I hit the cap once, after about 20 turns.
I pay for Pro and used Argon for about two weeks on long documents, code review and error logs. Where I quote other Korean users, it’s from Naver posts listed in Sources.
On this page
Is Gemini 4 Argon free?
No. Google’s API pricing page says Argon “is not available on the Free tier” (checked November 7, 2026), and when I logged out, it wasn’t in the Gemini app’s model picker. The only ways in are Pro, Ultra or paid API calls.
The free app plan runs Flash-Lite instead, per Google’s Korean plans page on October 30, 2026. Google AI Studio bills Argon through the same paid API, so it isn’t a free playground either.
Argon stopped after about 20 turns on Pro
Google doesn’t publish an Argon turn count. Its Gemini Apps limits page says app limits refresh every five hours until you reach a weekly cap, and that subscribers who hit a limit can keep going on Flash-Lite.
My one hit came on a weekday. I started a long document summary around 2 p.m. and kept at it for two or three hours, roughly 20 back-and-forth turns. Between 4 and 5 p.m., Argon stopped.
No dialog asked me to switch. A one-line notice about reaching the limit sat above the input box, and the model name had already changed to a smaller model. My next question went there without my choosing it.
Google’s help page says the notice tells you when the limit refreshes. Mine didn’t show a time. I finished two or three code questions on the smaller model and left the summary for later.
Argon came back between 7 and 8 p.m. I only noticed because I opened the picker after dinner and could select it again. That fits a five-hour window counted from when I started.
Since then I plan long Argon jobs for the morning or early afternoon. I summarize one document at a time and paste each result into a separate doc, so a cutoff doesn’t cost me finished work. I haven’t hit the limit again.
Other Pro users in Korea hit it too. One wrote that three or four uses finished the daily quota.
Checking your usage
On the web, open Settings at the bottom left, then Usage Limits. When I opened it a few days later, it showed roughly how much of my limit I’d used, not how many turns were left. It helps you pace a day, not plan one more long job.
Pro, Ultra and the plans that don’t include Argon
Pro and Ultra include Argon in the app; Plus doesn’t list it. These are Google’s Korean prices on October 30, 2026. Google sets plan prices by country, and its plans page showed me won only, so check the page for your own country before you compare.
| Plan | Monthly (Korea) | Argon in the app | Usage vs. free |
|---|---|---|---|
| Free | ₩0 | No (Flash-Lite) | Standard |
| Google AI Plus | ₩7,500 | Not listed | 2x |
| Google AI Pro | ₩29,000 | Yes | 4x |
| Google AI Ultra | ₩119,000 or ₩300,000 | Highest access | 5x or 20x Pro |
The multipliers come from Google’s limits page and are relative, so they don’t turn into an Argon turn count. I’m still on Pro at ₩29,000 a month, billed directly through Google One with a card rather than an app store, and I stayed on monthly billing instead of taking the annual discount.
- The student offer is a year of Google AI Plus, not Pro, so it doesn’t include Argon.
- Google’s Korean plans page listed no Pro free trial.
- Antigravity and Cursor have their own plans and quotas, separate from the Gemini app. Antigravity’s pricing page lists a $0 Individuals plan with basic weekly limits, but on November 7, 2026 it didn’t name Argon among its models.
$2 now, $4 from December 1: the API price
Through November 30, 2026, Argon costs $2 per 1M input tokens and $10 per 1M output tokens. From December 1 that doubles. Prices below are from Google’s pricing page, checked November 7, 2026.
| Rate per 1M tokens | Through Nov 30 | From Dec 1 |
|---|---|---|
| Input / output | $2 / $10 | $4 / $20 |
| Batch or Flex in / out | $1 / $5 | $2 / $10 |
| Priority in / out | $3.60 / $18.00 | $7.20 / $36.00 |
| Cached input | $0.10 | $0.20 |
For a sense of scale, take one test call with 10,000 input tokens and 2,000 output tokens, thinking included. That’s about $0.02 in plus $0.02 out, roughly $0.04 now and $0.08 from December 1. This is arithmetic from the table, not a bill I received.
Thinking tokens, the hidden reasoning before each answer, are billed at the output rate, so a short reply can still cost like a long one. The model code is gemini-4-argon, with a 1M-token context window. Only implicit caching applies.
The cheapest way to try Argon on the API:
- Create an API key in Google AI Studio on a project with billing on.
- Set a budget cap on that project first.
- In code that already calls Gemini, change only the model string to
gemini-4-argon, for exampleclient.models.generate_content(model="gemini-4-argon", contents=...)with the google-genai SDK. - Send a few short prompts and check the cost in the billing console before longer jobs.
My own API use was one personal project with a $10 spending cap, set in September 2026, where I ran only an evaluation config. When that cap trips, requests fail with a 403 right away.
Is Argon worth paying for over Gemini 3.1 Pro?
Only if heavy work comes up every week. Argon missed fewer middle clauses than 3.1 Pro on my long documents, but by my count it took two to three times as long to answer.
Its code review comments were also more specific, but you wait for them. Watching the screen, I saw long summaries take 25 to 45 seconds on Argon against 10 to 20 on 3.1 Pro, and review drafts take 20 to 35 against 8 to 15. I didn’t time them with a stopwatch, and the gap grew with longer documents.
So I route by task. One-line questions go to a lighter model, and Argon gets the long documents and reviews where a missed clause costs me.
If you mostly ask short things, like a grammar check or a single command, I wouldn’t subscribe for Argon. The free models handle that, and the extra wait gets annoying.
What I’m watching: December 1 and a free rollout
As of November 7, 2026, Google hasn’t announced Argon for the free plan; its launch post names only Pro and Ultra subscribers. None of the Korean posts showed a free user getting Argon.
Two things would change my answer. If Argon reaches the free app or Plus, I’d check whether Pro is still worth it for my weekly load. If you use the API, December 1 doubles your bill, and Batch at half price is the first thing I’d look at. Until then, Gemini 3.8 Flash has a free API tier for building and testing.
Sources
- Gemini API pricing, Google, checked November 7, 2026: no Argon free tier, Standard, Batch/Flex, Priority and cached rates before and after December 1, thinking tokens billed as output, Gemini 3.8 Flash free tier
- Gemini models, Google, checked November 7, 2026: model code and 1M-token context window
- Introducing Gemini 4 Argon, Google, October 13, 2026: release date and Pro/Ultra app access
- Gemini Apps usage limits, Google Help, checked October 30, 2026: five-hour refresh, weekly cap, Flash-Lite after the limit, plan multipliers, Usage Limits setting
- Google AI plans (Korea), Google, checked October 30, 2026: KRW prices, Flash-Lite on the free plan, no Pro trial
Show 3 more sources
- Google AI for students (Korea), Google, checked October 30, 2026: one-year Google AI Plus student plan
- Antigravity pricing, Google, checked November 7, 2026: $0 Individuals plan, no Argon named
- Naver searches (October 28, 2026; five Korean queries on Gemini 4 Argon access, limits and quality): first-hand posts from Argon users, including the quoted Pro user