New flagship model, big announcement, and then the catch in the fine print: you can't just try it. Welcome to Gemini 4 Argon.
Google made its new model generally available in mid-October. What that means for you depends almost entirely on how much you pay Google. Here is the sober version, without hype and without panic.
The developer forums show how touchy this is: many Pro subscribers were annoyed that they were left out at first. Understandable. Models like this cost Google a lot of computing power, so providers tier access by price. That goes for every big provider, not just Google.
What is Gemini 4 Argon?
A new AI model from Google, the "brain" behind Gemini. Google positions it as its currently strongest model. It is meant to help with multi-step work: researching, planning, writing code, checking results.
How good it really is will only show in independent tests. Vendor claims and benchmark tables (score lists providers use to compare models) are marketing with numbers attached. Useful, but not the last word.
Who can use it?
As things stand:
- Gemini app, free: No. The free offer stays with the smaller Flash models.
- Google AI Pro and Ultra: Yes. Argon was first meant for the top tier only and now reaches Pro subscribers in the Gemini app too.
- Google AI Plus: The cheaper plan loses the choice of Pro models from November. More in our post on Gemini's free tier.
- Antigravity (Google's coding tool): Pro subscribers get access, with their own weekly quota. The limit is built in.
- API (the programming interface for developers): Paid only. The free API tier does not include Argon.
Rule of thumb: free is enough for trying the Flash models. Argon costs money.
What does the API cost?
For developers, Google currently charges an introductory price: 2 dollars per million input tokens and 10 dollars per million output tokens. A token is roughly a syllable. A million of them is several novels.
At the current state, the introductory price runs until the end of November. After that it doubles, according to Google's pricing page. If you plan to build Argon into your own app, calculate with the higher price.
Is it worth it for you?
Depends on what you do.
You use AI now and then for emails, explanations, ideas? You don't need a flagship model. The free models handle that, with limits. For most beginners the upgrade is overkill.
You work with AI daily and hit limits? Then a trial of the Pro plan makes sense. Most providers let you cancel monthly, so you can test for a month and compare. Not a purchase recommendation, just a practical tip.
You develop yourself? Then the API is interesting. Test with small volumes before you extrapolate.
And the ecosystem matters more than the model name: if you already live in Google's world, with account, photos and mail, you get more from Gemini than someone who just wants a chat window.
What you can do now
- Open the Gemini app and check the model picker for what your account offers.
- Give a free model and a stronger one the same tricky everyday task. Compare for yourself.
- Before any subscription, check what you actually use. Cancel monthly means: try without regret.
Unpaid opinion, no sponsoring: the newest model is rarely the one you need. Usually the one before is enough, and that one is often free.