AI News
Written by Joshua Reiher
By Joshua Reiher
Some links on this page are affiliate links. If you sign up through them we may earn a commission, at no extra cost to you.
![]()
Short answer: for most people paying for Google AI Pro, Gemini 4 Neon is the one to use, because Gemini 4 Carbon is only in the Gemini app for Ultra subscribers. Neon is also cheaper to run per task. Carbon is still Google's strongest model for the hardest, long-running projects, so Ultra subscribers and developers with the toughest jobs may still prefer it.
Google released Neon on 18 November 2026 (US time), three weeks after it announced Carbon. Neon does not replace Carbon; Google positions it as a lower-cost option for complex work you do often. Here is how they compare, using only Google's own pages, read on 19 November 2026. For more about the app itself, see our Gemini listing.

Image: Google
| Gemini 4 Neon | Gemini 4 Carbon | |
|---|---|---|
| Announced | 18 November 2026 | 28 October 2026 (wider release 10 November) |
| Google's description | Balances depth and efficiency for complex, high-frequency work | For the hardest, long-horizon projects |
| Gemini app | Google AI Pro and Ultra | Google AI Ultra |
| Google AI Studio plans | AI Pro and Ultra | AI Ultra |
| Free plan | Not mentioned | Not mentioned |
| API price (per million tokens) | US$3 input, US$0.15 cached, US$15 output | US$2 input, US$10 output (introductory); US$4 input, US$20 output after that; cached input 95% off |
| Context window | 1 million tokens | Not stated in Google's posts |
| Best for | Coding, analysis and research you do often, on the Pro plan | The toughest long projects and security work, on Ultra |
Google says Neon brings "frontier-level intelligence with fast end-to-end task completion" to software engineering, knowledge work and science. It has four thinking levels and a 1 million token context window.
Its main selling point is cost. Google says that at the default "high" thinking level, Neon's cost per task is 32% lower than Carbon's, and that in the Gemini app this lets subscribers stretch their usage limits further.
Google's published results for Neon include 61 on the Artificial Analysis Intelligence Index at its highest thinking level (third overall, says Google), first place on the Artificial Analysis Coding Agent Index at 58.6, 65.2% on Terminal-Bench 4.1 and 70.5% on Terminal-Bench Science. In the Gemini app it is available to AI Pro and Ultra subscribers, and Tasks in Gemini can run on it.
Google calls Carbon "our strongest model with state-of-the-art cybersecurity and long horizon capabilities". It launched first to cyber defenders and partners, then on 10 November Google made it available through the API for paying developers and businesses and began rolling it out to all Google AI Ultra subscribers in the Gemini app, AI Mode in Search and Google Antigravity.
Google says Carbon ranked first on the Artificial Analysis Intelligence Index v5.0 with a score of 59 when it launched, and first on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench. Its API price is US$2 per million input tokens and US$10 per million output tokens for an introductory period, rising to US$4 and US$20.
Google compared the two directly on some security tests. It says Neon scored 90.5% on SEC-bench Pro (Carbon's launch post gave 88.2%), and that on ExploitGym Neon reached frontier-level performance using half the tokens of Carbon. On the Gray Swan prompt injection test, Google reports a 0.0% attack success rate for both. Note that some of these results were published at different times, and all are Google's own figures, not independent tests.
We used only Google's own pages, read on 19 November 2026: the Gemini 4 Neon announcement (https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/), the Gemini 4 Carbon launch post (https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbon/), the Carbon availability post (https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-carbon-now-available/) and Google's AI Studio plans page (https://aistudio.google.com/docs/google-ai-plans). We have not run our own tests. Image from Google's Neon announcement.
Not in the Gemini app at launch. Google lists it for Google AI Pro and Ultra subscribers and has not mentioned the free plan.
No. Google keeps Carbon for the hardest, long-running projects and positions Neon as a lower-cost model for complex work you do often.
Google calls Carbon its strongest model. On the tests Google shared, Neon is close to or ahead of Carbon in some areas, such as SEC-bench Pro, at a lower cost per task.
Already a member? The same buttons sign you in.
We'll email you a 6-digit code. No password needed. By continuing you agree to our Privacy Policy.
We sent a 6-digit code to . It expires in 10 minutes.
No email? Check your spam folder, or ·