On November 18, 2026, Google introduced Gemini 4 Neon, the third model in the Gemini 4 family in nine days. After Carbon (November 10) and Hydrogen (November 12), Neon sits in the middle: it reasons more deeply than Hydrogen, but finishes tasks faster and at a lower cost than Carbon. Google aims it at complex coding, analysis and business workflows.
Below you'll find prices, where to use it and the numbers Google reports. They are the company's own figures, not checked by third parties.
Where does Neon fit between Hydrogen and Carbon?
| Gemini 4 Hydrogen | Gemini 4 Neon | Gemini 4 Carbon | |
|---|---|---|---|
| Role according to Google | Fast everyday execution | Balance of depth and efficiency | Longest, hardest projects |
| Input / output (dollars per million tokens) | 1.50 / 7.50 | 3 / 15 | 2 / 10 until November 30, then 4 / 20 |
| Cached input | 0.075 | 0.15 | 0.10, then 0.20 |
| API free tier | Yes | No | No |
A token is a small piece of a word. Neon's context window is 1 million tokens: enough to hand it hundreds of pages at once. In Google's Antigravity development environment, Neon can also act as an "orchestrator": it splits the work and passes it to several copies of Hydrogen running in parallel.
How much does it cost and where can you use it?
- Gemini API and Google AI Studio, paid tier only (the model code is gemini-4-neon), plus Android Studio.
- Gemini app: for Google AI Pro and Ultra subscribers, from the model picker.
- Google Antigravity: for paid AI Pro and Ultra subscribers.
- Gemini for business: through the API and the model menu.
- Cybersecurity: access for people who defend systems through the Fairwind program, rolling out in the days after the announcement.
Read the price with a calendar in hand. Until November 30, Neon costs more than Carbon, which is on a promotional rate. From December 1, Carbon goes up to 4 and 20 dollars and Neon becomes the cheaper of the two.
What does Google claim in its tests?
Google reports a long list of results, measured by the company or by partners. We picked the ones closest to office work:
- Artificial Analysis index: 61 with reasoning at maximum ("xhigh"), third place behind larger, slower models.
- Zapier's AutomationBench (automations between apps): first, at 56.3%.
- Legal Research Bench and Tax Agent Bench (legal and tax research): first in both.
- Terminal-Bench 4.1: 65.2%. Hydrogen was at 57% in its own announcement.
- Cost per task: 32% lower than Carbon, with "high" reasoning.
What changes for people who work with AI?
If you have a Google AI Pro subscription, Neon gives you a more capable model than Hydrogen without paying for Ultra. It suits long requests: comparing contracts, building an analysis from several spreadsheets, checking a report before you send it. For quick questions, Hydrogen is still enough.
If you use the API to automate work, redo the math in December. Our take: if you're using Carbon today only because of the promotion, try Neon on the same tasks before December 1. If the results hold up, the switch saves a quarter on the price per token, and according to Google on the tokens used as well.
What does the announcement leave out?
- It doesn't mention Vertex AI, Google's cloud platform for companies: it only talks about Gemini for business.
- Most tests are internal or use a method the page doesn't describe; the full model card is a separate document.
- It doesn't say whether free Gemini app users will get Neon in some form, as happens with Hydrogen.
Frequently asked questions
Is Gemini 4 Neon free?
No. In the Gemini app you need an AI Pro or Ultra subscription, and the API pricing page excludes it from the free tier.
Does Neon replace Carbon?
No. Google keeps Carbon for the hardest, longest projects. Neon is for when you want good quality with less time and cost.
Which Gemini 4 model makes sense for office work?
For writing and summarizing, Hydrogen is enough. For long analyses across many documents, try Neon. You'll find the differences between the first two models in our article on Gemini 4 Hydrogen and Carbon.
Official sources
Transparency: this article is written with the help of artificial intelligence models, starting from the official sources listed, and is reviewed before publication.














