Skip to content
OriginalAI

Gemini 4 Neon is cheaper than Carbon — but only after December 1

Google's fourth Gemini 4 model in seven weeks sits between Hydrogen and Carbon at $3/$15 per million tokens. The catch: Carbon is on sale until November 30, so the cheaper model costs more for twelve more days.

Three translucent glass sheets standing on a pale windowsill in daylight: pale amber, a vivid green sheet in the middle, and smoky grey, the green sheet tinting the light that passes through it
The middle layer changes the light that passes through it — the middle model changes the price, once the calendar catches up.
On this page

In brief

  • Neon costs $3 per million input tokens and $15 per million output tokens — but Carbon is $2/$10 until November 30, so Neon is the pricier model for twelve more days.
  • From December 1, Carbon doubles to $4/$20 while Neon stays flat, making Neon 25% cheaper per token — Google's real pitch is cost per finished task, not per token.
  • Artificial Analysis scores Neon 61 on its Intelligence Index at xhigh, one point above Carbon — at $6.44 per task versus $7.76 for Carbon at standard pricing.
  • Neon is about twice as fast as Carbon in output speed (~110 vs ~57 tokens per second), finishing benchmark tasks in ~21 minutes instead of ~36.
  • Neon replaces Argon in the Gemini app's model picker and rolls out in the Gemini API, AI Studio, Antigravity and the Gemini app for Pro and Ultra subscribers.

Google announced its fourth Gemini 4 model in seven weeks on Tuesday, and the interesting part is not the model. It is the price tag, and the calendar.

Gemini 4 Neon sits in the middle of Google's lineup: heavier than the quick Hydrogen that arrived last week, lighter than the flagship Carbon. Tokens are the chunks of text a model reads and writes, and they are the unit the bill counts. Neon charges $3 per million tokens it reads and $15 per million it writes, with input it has already seen, and can reuse, billed at $0.15. Carbon, the model Neon is supposed to undercut, costs $2 and $10 right now. Neon's discount arrives on December 1, when Carbon's introductory price expires and its rates double to $4 and $20. Until then, the cheaper model is the more expensive one.

That is not a mistake. It is the whole story of Neon. Google is selling cost per finished task, not cost per token, and it is asking buyers to do the arithmetic on a rate card that changes in twelve days.

The family, in order

The Gemini 4 family now has four members, and they arrived fast. Argon came in October as the efficient frontier model. Carbon followed, announced October 28 and opened to all paying API developers on November 10. Hydrogen landed November 12 as the fast, cheap one, with a free API tier. Neon, announced November 18, is the fourth, adapted from Carbon and aimed at the work in between: long coding sessions, document-heavy analysis, and science workflows where depth and speed both count.

The price list, per million tokens on the paid tier of the Gemini API, reads like a ladder with one rung out of place. Hydrogen is $1.50 in and $7.50 out, and it is the only one with a free tier. Carbon and Argon are $2 and $10 through November 30, then $4 and $20 from December 1. Neon is $3 and $15, flat, with no introductory period and no free tier. Batch and Flex processing halve Neon's rates to $1.50 and $7.50.

Neon gets a million-token context window, the same as the rest of the family, and four thinking levels, low, medium, high, and xhigh, with high as the default. A thinking level is how hard the model works on a problem before answering: higher levels mean deeper reasoning, more tokens spent, and a bigger bill. The model reads text, images, audio, video, and PDFs, answers in text, and its knowledge runs to March 2026, according to Google's model documentation.

Two scoreboards that do not agree

Google's launch post leads with benchmarks, and the numbers are strong. The company says Neon ranks third overall on the Artificial Analysis Intelligence Index with a score of 61 at its xhigh thinking level, second on the Vals Index at 70.9 percent, and first on the Artificial Analysis Coding Agent Index at 58.6. On Terminal-Bench 4.1 it scores 65.2 percent, and on FrontierCode 2.0 it reaches 46.2 percent at its default high setting. The headline claim is about money: at high, Google says, Neon costs 32 percent less per task than Carbon.

Artificial Analysis, the independent benchmarking firm whose index Google cites, published its own Neon results the same day, and they tell a smaller story. Neon at xhigh scores 61, one point above Carbon at high, and third overall behind OpenAI's GPT-6.2 Astra and Anthropic's Claude Fable 5.2. The cost per task is $6.44, about 17 percent less than Carbon's $7.76, and the time per task is about 21 minutes against Carbon's 36. Neon writes roughly 15 percent more output tokens per task than Carbon, so the saving comes entirely from the lower token price, not from using fewer tokens. What it does buy is speed: about 110 output tokens per second, nearly double Carbon's 57.

These two accounts are not measuring the same thing. Google's 32 percent figure is at Neon's default high setting on Google's chosen tasks; Artificial Analysis's 17 percent is at xhigh on its own index. At high, the independent numbers are more generous to Neon, not less: Artificial Analysis has Neon matching Carbon's score of 60 at $3.74 per task, less than half of Carbon's $7.76. But there is a catch inside that catch. The $7.76 is Carbon priced at its standard $4 and $20 rates. At the discounted $2 and $10 that Carbon actually costs today, the same task would run about $3.88, by simple arithmetic on the published rates. Until November 30, Neon and Carbon cost roughly the same per finished task at the default setting, and Neon's real advantage is that it finishes sooner. On December 1, Carbon's task cost doubles and Neon's lead opens up. None of this makes either number wrong. It makes the date on the invoice part of the comparison.

Where you will find it

Neon is rolling out in the Gemini API on the paid tier in Google AI Studio, in Gemini for business, in Google Antigravity for paid Pro and Ultra subscribers, and in the Gemini app for Pro and Ultra subscribers. In the app, Neon takes the "Deeper analysis" slot in the model picker, and Argon moves out, according to 9to5Google's launch report. A separate version, Neon for cyber, is rolling out to trusted defenders through Google's Fairwind Program in the coming days.

The model itself is unremarkable in the best sense: a middleweight that does heavy jobs faster than the flagship, at a price that will be lower once the flagship's sale ends. The unusual part is the timing. Google shipped a model whose headline advantage does not exist yet, and will not until a discount expires on December 1. Anyone budgeting an API migration this month should price both dates, not one.

More from grep.tips

All articles