Have you ever put off trying a new tool because the price looked scary, only to find out it costs the same as the one you already use?
That is pretty much what happened yesterday. On November 10, Google opened Gemini 4 Carbon, its strongest AI model so far, to all paying developers through the Gemini API. Until then, only a small group of cybersecurity defenders had access through Google's Fairwind Program, after Carbon was first announced on October 28. Now anyone with a paid API key can call it in Google AI Studio and Android Studio, and businesses can use it through Gemini for business.
The catch is the price tag has a deadline. Until November 30, Carbon costs $2 per million input tokens and $10 per million output tokens, exactly the same as Gemini 4 Argon, the everyday model Google released a few weeks earlier. From December 1, both prices double to $4 and $20. So you have about three weeks to try Google's top model at its everyday-model price.
Argon is the daily driver, Carbon is the heavy lifter
Google is pretty clear about how the two models split the work. Argon is still the right pick for most coding and knowledge work. Carbon is built for the jobs that test the limits: complex, multi-step coding and debugging, long agentic workflows, and open-ended research.
Think of it like two cars in the same garage. Argon is the reliable sedan you drive every day. Carbon is the truck you rent when you need to move house.
One thing to keep in mind: Carbon burns more tokens by design, thinking longer and running extra verification loops. So even at the same per-token price, a Carbon job can cost more than an Argon job.
Photo by jonakoh _ on Unsplash
The proof is in the hard jobs
So what can Carbon actually do with all that extra thinking? Google points to a few examples.
On independent benchmarks, Carbon ranks first on CursorBench 4.0, Perplexity's WANDR, and Zapier's AutomationBench, and Google says it does so at more than two times lower cost per task than the next best outside model.
The example that caught my eye is Signal. Researchers used Carbon, alongside other models, to formally verify the code behind Signal's cryptographic protocol. Their agents generated millions of lines of code in Lean, a language for writing mathematical proofs, and proved key security properties in weeks. A job like that historically takes expert teams years. That is the kind of long, grinding task Carbon is made for.
Google is also previewing /swarm in Antigravity, its AI development platform, which turns one prompt into a coordinated team of agents you can watch in real time. Carbon takes the roles that need the deepest reasoning, like architectural design and auditing, while faster, lighter models handle the rest in parallel.
What should you actually do with this?
Here is my honest take. If you already pay for the Gemini API, try Carbon now, while the intro price lasts. Give it the job you have been avoiding: the messy multi-package migration, the bug that crosses three subsystems, the refactor you keep putting off.
But do your budgeting at the December prices, not the November ones. If a workflow only makes sense at $2 and $10, it will not make sense on December 1. And for everyday work, Argon is still the sensible default.
One more practical note: Carbon is not on the free tier of the Gemini API, and on the consumer side it is rolling out only to Google AI Ultra subscribers in the Gemini app, AI Mode in Search, and Antigravity. If you are on a lower plan, the API is the way in.
New models launch every week now. This one is worth a look for a simple reason: for three weeks, the strongest model costs the same as the everyday one.
So, what is the hardest job on your list right now? Maybe this is the week to finally hand it to something built for it.
