Google's Newest AI Model Is Off-Limits to Almost Everyone: Test the One You Can Use Before Its API Price Doubles December 1
Gemini 4 Carbon went to a small group of security defenders, with no date for businesses. Gemini 4 Argon is available now, and its API price doubles December 1. Here is the 20-minute test that tells you whether it belongs in your business.
What happened: On October 28, Google announced Gemini 4 Carbon. First access goes to a small group of cybersecurity defenders, with more trusted partners to follow. Google has not given a date for businesses or the general public.
The model you can use today is Gemini 4 Argon. It became generally available October 13, in the Gemini app on Google AI Pro and Ultra plans and through Google's API, the usage-billed connection software uses to reach the model.
The date that matters: Argon's API price goes from $2 to $4 per million input tokens and from $10 to $20 per million output tokens on December 1. A token is about three-quarters of a word.
So what for you: Skip the wait for Carbon. There is nothing to plan around until Google names a date. Spend 20 minutes finding out whether the model you can use beats what you already pay for.
The 20-minute test:
- Pick 3 real tasks from last week. A customer email, a contract summary, a spreadsheet question. Remove names and account numbers first.
- Run each task twice. Once in your current AI tool, once in Gemini with Argon selected. Use the same words both times. If you are not on a Google AI Pro or Ultra plan, a colleague or trial account can run it, or stay with your current tool and skip to step 4.
- Grade them blind. Paste the task and both answers into either tool with this:
Below is a task and two answers, labeled A and B. Grade each answer from 1 to 5 on three things: accuracy, whether I could use it with little editing, and whether it missed anything the task asked for. Quote the exact line behind any score of 2 or lower. List every fact in either answer that I should verify before using it. Then say which answer you would use and why, in two sentences.
TASK:
[paste task]
ANSWER A:
[paste first answer]
ANSWER B:
[paste second answer]
You make the final call. Read both answers yourself, because the tool doing the grading may favor its own style.
- Ask your vendors one question. If any software you pay for runs on Gemini, send this: "Does the December 1 Gemini 4 Argon API price change affect what we pay or how much usage we get?"
- Decide and write it down. Switch a task only if Argon wins it clearly. Otherwise keep what you have and move on.
What it's worth: If you or your developer call Argon directly, a job using 10 million input tokens and 2 million output tokens a month goes from $40 to $80 on December 1. Run your own numbers before then. If you only pay for a monthly Gemini plan, this API date is not your deadline.
Keep it safe: Before you paste any work material into a new tool, check its data settings and confirm your plan does not train on your chats. Keep client data in a private business workspace.