Free guide
Gemini 4 Neon: Which Thinking Level to Use (and How to Try It Today)
Google's new model scores about the same as its flagship on the default setting, in less than half the time. What it costs, what each of its four thinking levels gets you, and how to try it today.

The short version
- Neon sits between Carbon (Google's top model for the hardest work) and Hydrogen (the fast everyday one). It is rolling out in the Gemini API, Google AI Studio, the Gemini app for Google AI Pro and Ultra subscribers, Antigravity and Gemini for business.
- Price: $3 in and $15 out per million tokens, $0.15 for cached input, with a 1 million-token context window. There is no free API tier for Neon.
- Artificial Analysis scores by thinking level: low 44, medium 55, high 60, xhigh 61 (Carbon: 60). Going from high to xhigh roughly doubled the cost to run their index, for one more point. Google's own docs say start on high.
How it actually works
A neon sign on a dimmer
Think of Neon's thinking level as a dimmer. The brighter you set it, the longer the model thinks before it answers, and you pay for every one of those thinking tokens as output. Neon has four positions: low, medium, high and xhigh. If you don't set one, it runs on high.
What each level scored
Artificial Analysis ran its Intelligence Index on every level (checked November 19, 2026). Low scored 44, medium 55, high 60 and xhigh 61. Gemini 4 Carbon scores 60 on the same index. On high, Neon used about 40% fewer tokens than Carbon to run it (123 million output tokens vs 205 million) and averaged 14.8 minutes per task against Carbon's 37.1.
Why the top notch is rarely worth it
From high to xhigh the score went up one point while the cost to run the index went from about $3,695 to $7,788, roughly double, and the output tokens went from 123 million to 283 million. Google's thinking docs say the same thing in their own words: start with high, keep xhigh for hard multi-step problems, and note that xhigh can overthink simple prompts or run out of output tokens.
Per token it isn't the cheap one
Neon's $3 / $15 per million tokens is more than Carbon's current $2 / $10, a sale price Google lists through November 30, 2026 (Carbon goes to $4 / $20 on December 1). Neon saves by using fewer tokens and less time, not by a lower price per token, so measure your own jobs before you switch.
How to set it up
1.Get an API key
Open Google AI Studio's API keys page and create a key. Neon has no free tier in the API, so the key's project needs billing set up (the Set up billing button is on the same page).
https://aistudio.google.com/api-keys export GEMINI_API_KEY="YOUR_API_KEY"2.Call Neon on high
This is Google's own example from the thinking docs. high is the default, so you can also leave generation_config out.
curl -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \ -H "x-goog-api-key: $GEMINI_API_KEY" \ -H "Api-Revision: 2026-05-20" \ -H 'Content-Type: application/json' \ -d '{ "model": "gemini-4-neon", "input": "Audit this payment retry queue for race conditions.", "generation_config": {"thinking_level": "high"} }'3.Run one real job on each level and read the token count
Install the Python SDK, then run a task you already know the answer to on medium, high and xhigh. Compare the answer and the thinking and output tokens (thinking tokens are billed as output). Keep the lowest level that gets it right.
pip install -U google-genai from google import genai client = genai.Client() for level in ["medium", "high", "xhigh"]: it = client.interactions.create( model="gemini-4-neon", input="<your task>", generation_config={"thinking_level": level}, ) print(level, it.usage.total_thought_tokens, it.usage.total_output_tokens)4.No code? Use it in the Gemini app or AI Studio
Google AI Pro and Ultra subscribers can pick 4 Neon (labelled Deeper analysis) in the model drop-down in the Gemini app on mobile, desktop and web, in Chat and in Tasks. Google also says all Pro and Ultra subscribers can try it in AI Studio, and paid Pro and Ultra subscribers get it in Google Antigravity.
The fine print
- It is rolling out, so it may not show up for every account on day one.
- API prices per million tokens: $3.00 input, $0.15 cached input (implicit caching only), $15.00 output. Batch is half: $1.50 and $7.50. No free tier.
- The Artificial Analysis numbers are from their site on November 19, 2026 and can change as they re-run tests. Displayed scores are rounded (Neon on high is 59.6, Carbon 60.0).
- Artificial Analysis prices Carbon at its regular $4 / $20 in its cost totals, so cost comparisons with Carbon look different during Carbon's November sale. Compare tokens and time, or run your own jobs.
- xhigh output can be long: thinking tokens count toward max_output_tokens (65,536 by default in a normal call), so a low limit can cut an answer off.
FAQ
How much does Gemini 4 Neon cost?
$3 per million input tokens, $15 per million output tokens and $0.15 per million cached input tokens in the Gemini API, with a 1 million-token context window. Batch requests are half price. There is no free tier for Neon.
What are Gemini 4 Neon's thinking levels?
Four: low, medium, high and xhigh. The default is high. You set it with generation_config thinking_level in the API.
Is Gemini 4 Neon better than Gemini 4 Carbon?
On the Artificial Analysis Intelligence Index, Neon on its default high setting scores 60, the same displayed score as Carbon (59.6 vs 60.0), with about 40% fewer tokens and less than half the time per task; Neon on xhigh scores 61. Google still positions Carbon for its hardest, long-horizon projects. Test both on your own work.
Should I use xhigh?
Only for hard, multi-step problems. On Artificial Analysis' index, xhigh added one point over high and roughly doubled the cost, and Google's docs recommend starting with high.
Where can I use Gemini 4 Neon?
In the Gemini API (paid tier) through Google AI Studio and Android Studio, Gemini for business, the Gemini app for Google AI Pro and Ultra subscribers, and Google Antigravity for paid Pro and Ultra subscribers. Cursor also added it on launch day.
Sources
- Google: Introducing Gemini 4 Neon
- Gemini API: thinking levels
- Gemini API pricing
- Gemini 4 Neon model page
- Artificial Analysis: Gemini 4 Neon
- Sundar Pichai's launch post
November 19, 2026, against Google's launch blog, the Gemini API docs (models, pricing, thinking, quickstart) and the Artificial Analysis model page.
Want help building with tools like this?
Join the Shipping Skool community free. Paid members get live bootcamp calls every week and weekly-updated courses.
Join Shipping SkoolMore free guides
Chrome DevTools MCP: Let Your Coding Agent See the Page It Builds
Google's free, open-source way to give your coding agent a real Chrome window and the DevTools you already use. What it does, how it works, how to set it up, and the fine print.
GPT-6.2 Drops Its Reasoning When You Edit Old Messages: How to Keep It
Every reasoning item is now tied to the conversation before it. Edit that history and the API drops the reasoning that follows. Here's how it works and how to keep it.
Inkling 2 Mini: Run Thinking Machines' Looped AI Model on Your Mac
Thinking Machines' new open-weights model runs on a 32GB Mac with the wifi off. What it is, how the loop trick works, how to set it up, and the fine print.