
Key takeaways
- Google launched Gemini 4 Neon on Nov 18, 2026, and it replaces Argon in the Gemini app for AI Pro and AI Ultra subscribers.
- The app's model picker now has four entries: Auto, plus Everyday help for Hydrogen, Deeper analysis for Neon and Code & complex work for Carbon.
- Through the API, Neon costs $3 per million tokens in and $15 per million tokens out, which is 25% below what Argon and Carbon will cost after the Nov 30 promotion ends.
- Most of the benchmark numbers come from Google or the indexes it cites, and nothing here confirms they've been independently verified.
Google released Gemini 4 Neon on November 18, 2026, and it changes what you see when you open the model picker in the Gemini app. Neon sits in the middle of the Gemini 4 family. It’s live for Google AI Pro and AI Ultra subscribers, it takes over from Argon, and Gemini 3.1 Pro is being retired from the app.
The names Hydrogen, Neon, Argon and Carbon blur together fast, so I’d sort them by job.
What are the Gemini 4 models, and where did Argon go?
The Gemini 4 family has four models: Hydrogen for everyday help, Argon for code and analysis, Neon for complex, high-frequency work, and Carbon for the most demanding projects. In the Gemini app, Neon now replaces Argon. Google also describes Hydrogen as a replacement for Flash.
The picker labels are friendlier than the model names. According to 9to5Google, Hydrogen sits under Everyday help, Neon under Deeper analysis, and Carbon under Code & complex work. A fourth entry, Auto, chooses for you, and for paid users it runs on Gemini 3.5 Flash-Lite.
Argon’s future outside the app isn’t clear. Whether it stays in the API or other products hasn’t been reported.
Which model should you use for what?
For most paying subscribers, Neon is the sensible default for anything that takes real thought, Hydrogen handles quick everyday requests, and Carbon is for the hardest long projects. That follows Google’s own positioning, though the task matching below is my judgment, not Google’s.
| What you’re doing | Picker choice | Why |
|---|---|---|
| Quick questions, short drafts, everyday lookups | Everyday help (Hydrogen) | Google pitches it for fast everyday tasks. |
| Reading a spreadsheet, PDF or photo and pulling out answers | Deeper analysis (Neon) | Google says Neon reads these files more precisely than before. |
| Multi-step Tasks, like checking billing statements against receipts | Deeper analysis (Neon) | Google says Tasks on Neon finish faster and stretch usage limits further. |
| Research that needs Gmail, Docs and the web together | Deeper analysis (Neon) | Google says Neon draws on all three. |
| Long-horizon projects and the hardest coding work | Code & complex work (Carbon) | Google’s pick for the most complex jobs. |
| You don’t want to choose | Auto | It adapts to the request. |
If you’re deciding between models at another company’s lineup, we walked through a similar choice in our guide to picking between GPT-6.2 Astra and Sol. The logic carries over: use the cheapest model that does the job, and step up only when the answer isn’t good enough.
What does Gemini 4 Neon actually do differently?
Google describes Neon as a frontier-level model built for complex, high-frequency workflows, aimed at being faster and cheaper than larger models. It has a 1 million-token context window and four thinking levels: Low, Medium, High and Extra High.
The consumer-facing claims are modest and practical. Google says Neon writes more clearly, reads uploaded spreadsheets, PDFs and photos more precisely, and pulls details from Gmail and Docs while using web sources for current research. Its own examples for Tasks in the Gemini app are analyzing billing statements and cross-checking receipts to negotiate a refund.
Paid AI Pro and Ultra subscribers can also use Neon in Google Antigravity. Whether the app lets you pick a thinking level isn’t spelled out in the coverage.

Is Neon better than Carbon?
Not across the board, and nobody has shown that it is. 9to5Google says Neon performs comparably to Carbon, and possibly better, on coding, reasoning and agentic tasks, at Flash-level speed and lower cost. The same report ranks Neon third, just behind larger, slower models.
Neowin’s headline says Google claims Neon is three times faster than Carbon and can beat it on some benchmarks. Only the headline was available, so how the speed was measured is unknown.
The scores look strong on paper. 9to5Google’s rundown gives Neon 61 on the Intelligence Index from Artificial Analysis with thinking set to extra high, good for third place. On Arena’s Agent Arena Index it scores 1336, second globally across more than 4,200 real-world agent sessions. It posts 58.6 on the Artificial Analysis Coding Agent Index, which is first. The list also credits it with the top spot on Terminal-Bench Science at 70.5%, RiemannBench at 84.0% and HLE-Diamond at 71.3%.
These figures appear to be published or promoted by Google, so treat them as Google’s claims rather than verified results. A model that tops a benchmark isn’t automatically better at your tax questions.
How much does Neon cost?
In the app, Neon comes with your AI Pro or AI Ultra plan, so there’s no per-use price. The per-token price matters if you build on the API: Neon runs $3 per million tokens in and $15 per million tokens out, twice Hydrogen’s price.
The comparison with Argon and Carbon is odd right now. Until November 30, those two sit at promotional pricing of $2 input and $10 output, which makes Neon the more expensive option for the moment. After that date they rise to $4 and $20, and Neon becomes 25% cheaper. Neowin’s headline also calls the current pricing situation “a little unusual.”
Android Authority adds that cached input tokens come with large discounts, but gives no figures. Hydrogen’s actual price wasn’t reported either.
Why this matters if you pay for AI Pro
The biggest change is access. Android Authority says Carbon was restricted to AI Ultra, while Neon works on AI Pro plans, so it lowers the barrier to entry. 9to5Google doesn’t make the Ultra-only claim, so check your own plan’s picker before assuming anything.
In practice, I’d start with Neon for anything involving your files, email or multi-step Tasks, and drop to Hydrogen for quick chat. Reach for Carbon only if Neon’s answer falls short on a big project. Switching costs you nothing but a tap, so test one stubborn task on two models before settling on a habit.
Better prompts help more than a bigger model, too. Our guide to writing better AI prompts covers that.
Android Authority also reads the launch as Google matching the multi-model approach of Anthropic and OpenAI, with a model for each rung of cost and difficulty. That’s a fair description of what the picker now looks like.

What Google hasn’t said yet
The list of gaps is long. Usage limits for Neon versus Carbon and Hydrogen on AI Pro and AI Ultra haven’t been reported, which is the number most subscribers will care about.
Also unknown:
- Whether the speed claim holds, and under which conditions (tokens per second, latency, thinking level).
- Whether the benchmark results have been independently confirmed by Artificial Analysis, Arena, Surge AI or others.
- Why Neon is priced above promotional Carbon, and whether Carbon’s $4/$20 post-promo price will hold.
- How Neon compares with rival Anthropic and OpenAI models on price and benchmarks.
A list of related coverage also points to a GitHub post on Neon in GitHub Copilot. What that integration includes, and who gets it, is unconfirmed.
Related gear
Disclosure: some links in this article are affiliate links or our editor's own courses and book. If you buy through them, TechyCluster or its editor may earn money at no extra cost to you. This does not influence what we cover or how we cover it. As an Amazon Associate, TechyCluster earns from qualifying purchases.
Frequently asked questions
What is Gemini 4 Neon?
It's a mid-tier model in Google's Gemini 4 family, launched Nov 18, 2026. Google describes it as a frontier-level model for complex, high-frequency work, positioned between Hydrogen and Carbon.
Is Gemini 4 Neon available on the free Gemini plan?
Reports say it's live for AI Pro and AI Ultra subscribers. Whether free users get any access hasn't been reported.
What happened to Gemini 4 Argon?
Neon replaces Argon in the Gemini app, and Gemini 3.1 Pro is being retired from the app. Whether Argon stays in the API or other products hasn't been reported.
How much does Gemini 4 Neon cost per million tokens?
API use costs $3 per million tokens in and $15 per million tokens out, twice Hydrogen's price. After Nov 30 that is 25% cheaper than Argon and Carbon, but it's higher than their promotional $2/$10 until then.
Get the next one in your inbox. One email a day with the tech stories that matter, explained in plain English. Subscribe free.
Sources
This article was compiled from reporting by the following outlets. Links go to the original reports.
- 9to5Google: Google launches Gemini 4 Neon with Carbon quality at Flash speeds, live for AI Pro
- Neowin: Google launches Gemini 4 Neon, claims it's three times faster than Carbon
- Android Authority: Gemini 4 Neon is ready to get some work done for AI Pro power users
- 9to5Google: Google launches Gemini 4 Neon with Carbon quality at Flash speeds, live for AI Pro
