By Robert Shakir, AI & Digital Marketing Consultant · Published November 19, 2026 · Facts checked against Google’s announcement on November 19, 2026
Google released Gemini 4 Neon on November 18, 2026. It’s a new middle model in the Gemini 4 family, and the part most people will notice is simple: you no longer need the top AI Ultra plan to get near-Carbon reasoning. Neon is live for Google AI Pro subscribers in the Gemini app today. My quick read for most paying users: switch your default to Neon for research, documents and spreadsheets, and keep Hydrogen for quick everyday questions.
This is a news analysis based on Google’s announcement and press coverage. I haven’t run my own benchmarks on Neon yet, so the performance figures below are Google’s claims.
What Google actually announced
Neon sits between Hydrogen (the fast, cheap everyday model) and Carbon (the largest one). Google says it’s aimed at complex coding, analysis and business workflows, and that it finishes tasks faster and at a lower cost than its bigger sibling.
The key facts from Google’s post:
- Price: $3 per million input tokens, $0.15 per million cached input tokens, and $15 per million output tokens.
- Context window: 1 million tokens, so it can take in long PDFs, spreadsheets, audio and video in one go.
- Four thinking levels (low, medium, high and extra high), which let you trade answer quality against speed and token spend.
- Where it’s available: the Gemini API in Google AI Studio, Gemini for business, Google Antigravity for paid Pro and Ultra users, and the Gemini app model picker for Pro and Ultra subscribers.
- Benchmarks Google highlights: #3 overall on the Artificial Analysis Intelligence Index (score 61 at extra high), #1 on that firm’s Coding Agent Index (58.6), and #1 on Zapier’s AutomationBench (56.3%).
A security-focused version, “Neon for cyber,” goes to approved defenders through Google’s Fairwind Program over the coming days. That one isn’t for regular users.
The new Gemini model picker
Neon also cleans up a confusing lineup. According to 9to5Google, it replaces Gemini 4 Argon, and Gemini 3.1 Pro leaves the app. Paid users now see four choices:
| Picker label | Model | Best for |
|---|---|---|
| Auto | Picks for you | People who don’t want to think about it |
| Everyday help | Gemini 4 Hydrogen | Quick answers, rewrites, short emails |
| Deeper analysis | Gemini 4 Neon | Reports, spreadsheets, research, most coding |
| Code & complex work | Gemini 4 Carbon | The hardest, longest projects |
The big change for buyers: Carbon needs AI Ultra, but Neon works on AI Pro too, as Android Authority points out. That lowers the entry price for serious analysis work by a whole plan tier.
API prices compared: the math for developers
If you build on the API, the numbers matter more than the benchmarks. Here’s how Neon stacks up against the rest of the family, using the rates Google and 9to5Google published on November 18:
| Model | Input (per 1M tokens) | Output (per 1M tokens) | Example job cost* |
|---|---|---|---|
| Gemini 4 Hydrogen | $1.50 | $7.50 | $15 |
| Gemini 4 Neon | $3 | $15 | $30 |
| Argon / Carbon (promo until Nov 30) | $2 | $10 | $20 |
| Argon / Carbon (from Dec 1) | $4 | $20 | $40 |
*Example job: 1,000 requests, each with 5,000 input tokens and 1,000 output tokens (5M in, 1M out), with no caching. My own illustration, not a Google figure.
Two things jump out. First, until November 30 Carbon is actually cheaper per token than Neon, because of its launch promo. Google says Neon still costs about 32% less per task than Carbon at the default setting, since it uses fewer tokens to finish a job, but per-token, the promo wins for now. Second, the cached-input rate is a 95% discount. If your app sends the same long instructions or documents every time, set up caching before you compare prices. It changes the bill more than the model choice does.
Who should switch, and who can wait
Switch now if you pay for AI Pro and use Gemini for long documents, spreadsheet analysis or research briefs. You get a much stronger model at no extra cost. Just pick “Deeper analysis” in the model menu.
Developers on Argon should test Neon soon, since Argon is being replaced. Run your own prompts side by side before December 1, when Carbon’s promo price ends.
Wait if you mostly ask short questions or write quick messages. Hydrogen is half the price on the API and plenty for that work. And if you’re on the free plan, nothing changes for you today.
Small businesses comparing ChatGPT and Claude: Google is clearly pitching Neon at the same buyers as OpenAI’s and Anthropic’s mid-tier models. Benchmarks from the vendor are a starting point, not proof. See how the three assistants differ in daily use in my ChatGPT vs Claude vs Gemini comparison.
What I’m unsure about
Google doesn’t say how many Neon messages Pro users get before hitting limits, only that it “stretches usage limits further” than larger models. Until that’s clear, heavy users should watch for caps. The benchmark rankings also come from a launch post, and independent tests over the next few weeks will show how well they hold up.
Robert’s take
In my consulting work, the teams that get value from AI aren’t the ones on the most expensive plan. They’re the ones who match the model to the job. Neon makes that easier on AI Pro: use it for the reports and spreadsheets that eat your afternoon, and don’t waste it on two-line emails. If you’re building on the API, sort out caching first. That’s where the real savings are.
Running AI locally instead? A laptop with a strong NPU helps. Here’s what an NPU actually does and my picks for the best AI laptops of 2026.
Sources
- Google, Introducing Gemini 4 Neon (Nov 18, 2026)
- 9to5Google, Google launches Gemini 4 Neon with Carbon quality at Flash speeds (Nov 18, 2026)
- Android Authority, Gemini 4 Neon is ready to get some work done for AI Pro power users (Nov 18, 2026)
Written by Robert Shakir, AI & Digital Marketing Consultant. See the editorial policy.