Menu

Gemini 4 Neon: Google's new $3 AI model lands between Hydrogen and Carbon. Should you switch?

Comingwave team · 7 minute read · published

Comingwave logo and tagline

Gemini 4 Neon is Google's new middle model, released on 18 November 2026 for developers on the paid Gemini API and for Google AI Pro and Ultra subscribers in the Gemini app. It costs $3 per million input tokens and $15 per million output tokens, sits between the fast Gemini 4 Hydrogen and the heavyweight Gemini 4 Carbon, and Google pitches it at the spreadsheet, PDF and coding work that small and medium businesses already hand to AI. Here is what changed and what to test before you switch.

What happened

Google announced the model in "Introducing Gemini 4 Neon", a post by Koray Kavukcuoglu, SVP of Google DeepMind and Google's Chief AI Architect, published on 18 November 2026. Google describes Neon as "bringing frontier-level intelligence with fast end-to-end task completion across complex software engineering, knowledge work, and science workflows".

The Gemini API changelog entry for the same day lists the model ID gemini-4-neon as generally available, which means it is a finished product rather than a preview. It is on the paid tier only. The same entry carries a second change that matters to office work: developers can now send Microsoft Office and OpenDocument files (Word, PowerPoint and Excel) straight to the Gemini API, and pick out only the pages, worksheet tabs or cell ranges they want processed and paid for.

Neon is the latest step in a fast run of releases. The Decoder reported on 18 November 2026 that it is Google's fourth new Gemini 4 model in four weeks. We covered two of the earlier ones: Gemini 4 Carbon's general release and Gemini 4 Hydrogen, the fast everyday model.

Key details

ItemWhat Google says
Model IDgemini-4-neon, generally available on the paid Gemini API tier (not the free tier)
Price$3 per million input tokens, $0.15 per million cached input tokens (a 95% discount), $15 per million output tokens
Context window1 million tokens shared by input, thinking and output
Thinking levelslow, medium, high (the default) and xhigh
Where it runsGemini API in Google AI Studio, Gemini for business (API and model dropdown), Google Antigravity for paid Pro and Ultra subscribers, the Gemini app for Pro and Ultra subscribers
Position in the familyBetween Hydrogen (fast everyday work) and Carbon (the hardest, long-horizon projects)
Knowledge cutoffMarch 2026, according to the model card

A token is a small chunk of text, roughly part of a word; AI services charge for the tokens they read and the tokens they write. Thinking tokens, the model's working before it answers, are billed as output.

Why it matters

The interesting part is the price-for-quality position. Google says Neon ranks third overall with a score of 61 on the Artificial Analysis Intelligence Index at its xhigh setting, second on the Vals Index at 70.9%, and first on both the Legal Research Bench and Tax Agent Bench. These are Google's own figures about outside benchmarks, so treat them as a reason to test, not as proof that it suits your work.

On cost, Google says that at the default high setting Neon has a 32% lower cost per task than Carbon, and roughly 50% to 80% lower cost per task than top competing frontier models on the Vals Index and Agent Arena. Cost per task matters more than the price per token: a model that thinks less and finishes in fewer steps can cost less overall even at a higher token rate.

The calendar complicates the comparison. Carbon is on an introductory price of $2 per million input tokens and $10 per million output tokens until 30 November 2026, then $4 and $20. VentureBeat noted on 18 November 2026 that Carbon is cheaper per token than Neon through November, but from 1 December Neon's rates are 25% below Carbon's. Hydrogen costs $1.50 and $7.50, half Neon's price.

Google also leans on behaviour, not only scores. It says Neon recorded its lowest fabrication rate yet in internal testing, took shortcuts on 2.45% of tasks in its coding cheating test, and, with real-time safeguards at its highest thinking level, resisted every attack in Gray Swan's indirect prompt injection benchmark. Prompt injection is when hidden instructions in a web page, email or document try to hijack an AI agent. The Gemini 4 Neon model card is more cautious: it says the model may still hallucinate, meaning it can state things that are not true, and that the full Safety and Security Report is still to come.

What this means for businesses

Most small businesses will meet Neon in one of two ways: inside the Gemini app or Gemini for business, or inside software a developer has built on the Gemini API. The questions differ for each.

If your team uses the Gemini app, Google's examples sound like ordinary office jobs: reading spreadsheets, PDFs and photos more precisely, turning dense material into reports, pulling details from Gmail and Docs, analysing billing statements to see why a utility charge spiked, and cross-checking receipts and documents for a refund claim. Those are jobs where a person can check the answer against the original, which is the right way to use any model. A bookkeeper should still trace each figure back to the source document. A recruitment office using it to summarise CVs should confirm work history separately; a verification platform such as Talent Tabloid helps there. Our page on work for professional services firms covers where these tools fit in a practice.

If you pay for software built on the Gemini API, the change is quieter but more important. Google's "What's new in Gemini 4 Neon" guide says a switch from another Gemini 4 model needs no request changes beyond a new model ID, so a developer can move your app quickly. It also says hard prompts can take several minutes, that client timeouts should be at least 15 minutes, and that every token, thinking included, is billed. A switch that looks trivial in the code can change both speed and the monthly bill.

The Office file change helps businesses that live in Excel and Word. Sending only the worksheet tab or page range you need, rather than the whole file, cuts what you pay for and limits how much of your data goes to the model. This is the kind of connection we build in business systems and integrations work.

A short checklist before you change anything:

  • Ask your developer or software vendor which Gemini model your tools call today, and whether they plan to move to Neon.
  • Run the same set of real tasks through your current model and Neon at high, and compare quality, time and the token bill, not the benchmark tables.
  • Put Carbon's price change after 30 November 2026 in your budget if any of your tools use Carbon.
  • Set spending alerts on your Gemini API project before testing, because thinking tokens are billed as output.
  • Check which Google terms cover the account your staff use before they upload customer records, and keep a person checking any figure that goes to a client.
  • Limit what an AI agent can open and send; our guide to AI agent permissions, data access and cost walks through it.

Choosing between Hydrogen, Neon and Carbon is now a cost decision as much as a quality one. If you want help testing models against your own work or wiring them into your systems, ask us for a quote: the first consultation is free and you get a written quote before any work starts.

Frequently asked questions

What is Gemini 4 Neon?

Gemini 4 Neon is a Google AI model released on 18 November 2026. Google places it between Gemini 4 Hydrogen, built for fast everyday work, and Gemini 4 Carbon, built for the hardest long projects, and aims it at complex coding, analysis and office work.

How much does Gemini 4 Neon cost?

On the paid Gemini API, Google charges $3 per million input tokens, $0.15 per million cached input tokens and $15 per million output tokens, with thinking tokens billed as output. In the Gemini app it comes with Google AI Pro and Ultra subscriptions.

Is Gemini 4 Neon available on the free tier?

No. Google's developer guide says Neon is on the paid tier of the Gemini API only. In the Gemini app it is offered to Google AI Pro and Ultra subscribers, and businesses can reach it through Gemini for business.

Should we switch from Gemini 4 Carbon or Hydrogen to Neon?

Test first. Neon is pitched as close to Carbon on quality at a lower cost per task, while Hydrogen is half Neon's price. Run your own real tasks through each and compare results, time and cost before you change anything.

Can Gemini 4 Neon still make mistakes?

Yes. Google reports its lowest fabrication rate yet in internal testing, but the model card still says Neon may hallucinate. Keep a person checking figures, legal points and anything sent to customers.

Key takeaways

  • Google released Gemini 4 Neon on 18 November 2026, between Hydrogen and Carbon.
  • It costs $3 per million input tokens and $15 per million output tokens, paid tier only.
  • Carbon's introductory price ends on 30 November 2026, which changes the comparison.
  • The Gemini API now reads Word, PowerPoint and Excel files directly, by page, tab or range.
  • Test Neon on your own tasks, set spending alerts and keep a person checking the output.

Need help with your business technology?

Tell us what you need. We aim to reply quickly.

Get a free quote