---
title: "Gemini 4 Neon Is Out, and It Undercuts Carbon From December"
description: "Google released Gemini 4 Neon for paid Gemini API users and AI Pro and Ultra subscribers, at $3/$15 per million API tokens. When it beats Carbon on price."
url: https://publishmyblog.com/ai/google-releases-gemini-4-neon/
section: "AI"
author: "Saeed Shahriar"
published: 2026-11-19T13:18:10.304Z
---

# Gemini 4 Neon Is Out, and It Undercuts Carbon From December

Google's new model between Carbon and Hydrogen costs more than Carbon until November 30 and less from December 1, and it isn't on the API's free tier.

Google released Gemini 4 Neon on 18 November, a new model that it places between Carbon, its pick for the hardest long projects, and Hydrogen, the fast one for everyday work.

On Google AI Pro, AI Ultra, the paid Gemini API or Gemini for business? Then Gemini 4 Neon is rolling out to you from today.

You pay **$3 per million input tokens and $15 per million output tokens**, with cached input at $0.15, says [Google's Gemini 4 Neon announcement](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/). The context window is 1 million tokens.

So is it cheaper than Carbon for you? That depends on which week you ask, because Carbon's own price is about to change.

## Gemini 4 Neon vs Carbon and Hydrogen on price

Carbon sits on a launch price of $2 and $10 per million tokens through November 30, then $4 and $20 from December 1, as we covered when [Gemini 4 Carbon came out](https://publishmyblog.com/ai/google-releases-gemini-4-carbon/).

**This month Neon costs you about 50% more per token than Carbon, and from December 1 about 25% less.**

Here are the four Gemini 4 API prices side by side, per million tokens, from Google's pricing page as of 19 November 2026:

| Model | Input | Output |
|---|---|---|
| Gemini 4 Hydrogen | $1.50 | $7.50 |
| Gemini 4 Neon | $3 | $15 |
| Gemini 4 Carbon, to November 30 | $2 | $10 |
| Gemini 4 Carbon, from December 1 | $4 | $20 |
| Gemini 4 Argon, to November 30 | $2 | $10 |
| Gemini 4 Argon, from December 1 | $4 | $20 |

Gemini 4 Argon, out since October 13, has the same prices as Carbon, so Neon costs you about 50% more per token than Argon this month, and from December 1 about 25% less.

But your bill isn't only the price per token.

A model that thinks longer, or writes longer answers, burns through more tokens on every job you give it, and on Neon those thinking tokens bill as output.

Google says **Neon's cost per task at its default high thinking level is 32% lower than Carbon's**.

[The Decoder's read of the Artificial Analysis numbers](https://the-decoder.com/googles-latest-ai-model-is-faster-cheaper-and-smarter-than-its-flagship-predecessor-from-just-weeks-ago/) shows a smaller gap at xhigh. A task there cost about $6.44 on Neon against $7.76 on Carbon at full price.

Roughly 17% less.

**Neon used about 15% more output tokens per task**, so the whole saving comes from its lower list price.

## It's quicker than Carbon

Meanwhile, The Decoder reports from the Artificial Analysis results that Neon writes about 110 tokens a second, where Carbon manages 57.

A benchmark run took Carbon about 36 minutes, The Decoder says. Neon needed about 21.

That matters to you if **your agents or scripts sit waiting on every answer** before the next step.

## Where you can use it

Here's where Google is rolling it out:

- The Gemini API paid tier, in Google AI Studio and Android Studio
- Gemini for business, through the API and its model dropdown
- The Gemini app, for Google AI Pro and Ultra subscribers
- Google Antigravity, for paid AI Pro and Ultra subscribers
- Neon for cyber, a security version for trusted defenders in Google's Fairwind Program, in the coming days

Do you hand off tasks in the Gemini app?

Google says **those tasks can now run on Neon**, which it says gets through complex, multi-step jobs faster than the larger models and stretches your usage limits further.

## How good is it?

Google says Neon scores 61 on the Artificial Analysis Intelligence Index at its xhigh setting, which puts it third overall, behind what Google calls larger, slower models.

The Decoder puts Carbon at 60.

Who's ahead? OpenAI's GPT-6.2 Astra and Anthropic's Claude Fable 5.2, The Decoder says, and it points out that **Google's own comparison table leaves both of them out**.

For coding, Google puts Neon first on the Artificial Analysis Coding Agent Index, at 58.6.

Those are Google's and Artificial Analysis's numbers.

## What we checked

Google's Gemini API pages (read 19 November) add two things the announcement leaves out, and confirm a third.

[The API pricing page](https://aistudio.google.com/docs/pricing) says Gemini 4 Neon is **not available on the free tier**. Thinking tokens bill at $15 per million output tokens, and caching is implicit only.

[Neon's model page](https://aistudio.google.com/docs/models/gemini-4-neon) lists text as its only output: no image or audio generation, no Live API. Max output is 65,536 tokens by default, and you need long decode to reach 1,048,576.

[The API release notes](https://aistudio.google.com/docs/changelog) mark it generally available on 18 November, with four thinking levels: low, medium, high (the default) and xhigh.

## Why it matters

Running Carbon in the API today? Test Neon before December 1. That's when Carbon's price doubles.

If you send a few big jobs a day, the gap is small, but if your site runs a lot of calls through Carbon each day, that one date can move your bill a lot.

If you build on the free tier, though, Neon won't run for you. **Stay on Hydrogen or budget for the paid tier.** Google also made Hydrogen [the default model in AI Mode](https://publishmyblog.com/ai/gemini-4-hydrogen-ai-mode-default/).

What I'd do this week, depending on where you use Gemini:

- On Carbon in the API: rerun ten of your real prompts on Neon at high and compare the cost per task
- On the free tier: stay on Hydrogen
- On AI Pro in the Gemini app: try Neon on one long PDF or spreadsheet you already know well

My read: for you as a site owner or marketer, the bigger change is in the Gemini app, where I expect most of you will meet Neon long before you ever touch its API.

Google pitches Neon at spreadsheets, PDFs, charts and a complex brief. **If you're on AI Pro, your plan already includes it.**

Still, The Decoder counts four new models in seven weeks, so I'd pick by your own bill and your own test prompts rather than by the charts Google chose to show.

That's a lot of models.

## Common Questions

**How much does Gemini 4 Neon cost in the Gemini API?**

$3 per million input tokens and $15 per million output tokens, with cached input at $0.15. Thinking tokens count as output, and Gemini 4 Neon isn't on the API's free tier.

**Is Gemini 4 Neon cheaper than Gemini 4 Carbon?**

Only from December 1. Carbon costs $2 and $10 per million tokens through November 30, then $4 and $20, which leaves Gemini 4 Neon 25% cheaper per token.

**Can I use Gemini 4 Neon in the Gemini app?**

Yes, with a Google AI Pro or Ultra subscription. Paid AI Pro and Ultra subscribers also get Gemini 4 Neon in Google Antigravity.

## Sources

- [Introducing Gemini 4 Neon](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/), Google
- [Google's latest AI model is faster, cheaper and smarter than its flagship predecessor from just weeks ago](https://the-decoder.com/googles-latest-ai-model-is-faster-cheaper-and-smarter-than-its-flagship-predecessor-from-just-weeks-ago/), The Decoder
- [Gemini Developer API pricing](https://aistudio.google.com/docs/pricing), Google AI Studio
- [Gemini 4 Neon](https://aistudio.google.com/docs/models/gemini-4-neon), Google AI Studio
- [Release notes](https://aistudio.google.com/docs/changelog), Google AI Studio
