Google released Gemini 4 Neon on November 18. It takes Argon's place in the Gemini app for AI Pro and AI Ultra subscribers, runs in Antigravity, and costs $3 and $15 per million tokens on the paid Gemini API. Google says it matches Carbon on coding and agent work for less. Who gets it and what to do.
If you pay for Google AI Pro, the model menu in the Gemini app changes this week. Gemini 4 Neon, released November 18, is rolling out to AI Pro and AI Ultra subscribers and takes the slot Gemini 4 Argon held. Developers can call it on the paid Gemini API for $3 per million input tokens and $15 per million output tokens. Google says it matches its top model, Gemini 4 Carbon, on coding and agent work while finishing tasks faster and for less.
What Google released
Neon is the third Gemini 4 model in a week. Carbon reached the paid API and AI Ultra on November 10, and the fast Gemini 4 Hydrogen followed on November 12. Google DeepMind’s Koray Kavukcuoglu, who wrote the announcement, places Neon between the two: Carbon for long, hard projects, Hydrogen for quick everyday work, and Neon to balance depth and speed.
9to5Google reports that Neon replaces Argon and that Gemini 3.1 Pro is going away. Paid users get a new model menu with four choices:
- Auto, which picks a model for each prompt.
- Everyday help, Gemini 4 Hydrogen.
- Deeper analysis, Gemini 4 Neon.
- Code and complex work, Gemini 4 Carbon.
Who gets it
According to Google’s post, Neon is rolling out starting November 18 to:
- Gemini app. AI Pro and AI Ultra subscribers, from the model menu.
- Google Antigravity. Paid AI Pro and AI Ultra subscribers.
- Developers. The paid tier of the Gemini API in Google AI Studio and Android Studio. The model ID is
gemini-4-neon. - Work accounts. Gemini for business, through the API, the model menu and Antigravity.
Google’s post doesn’t mention free users or AI Plus. A security version, Gemini 4 Neon for cyber, goes to vetted defenders in Google’s Fairwind Program in the coming days.
What it costs
On the API, Neon costs $3 per million input tokens, $0.15 per million cached input tokens and $15 per million output tokens. It has a 1 million token context window and four thinking levels, with high as the default.
That is twice Hydrogen’s price. Today it is also more than Argon and Carbon, which cost $2 and $10 per million tokens on an introductory price through November 30. After that, both rise to $4 and $20, and Neon becomes the cheaper option. Google also says Neon’s cost per task at the default setting is 32% lower than Carbon’s, because it finishes jobs with less work.
In the app, nothing changes on your bill. Neon comes with the AI Pro and AI Ultra plans you already pay for. Our AI plans comparison lists every Google tier.
What Google says it can do
These are the results Google cites in its post, from its own tests and outside leaderboards. Independent hands-on tests are not out yet.
- Overall. 61 on the Artificial Analysis Intelligence Index at the highest thinking level, third overall, and first on its Coding Agent Index at 58.6.
- Office work. First on Vals AI’s legal research and tax benchmarks, and first on Zapier’s AutomationBench at 56.3%.
- Coding. 65.2% on Terminal-Bench 4.1.
- Security. 90.5% on SEC-bench Pro, up from the 88.2% Google reported for Carbon.
- Behavior. Google says Neon makes things up less often than any model it has tested in Antigravity.
Google also says Neon is better at reading spreadsheets, PDFs and photos you upload, and at pulling details from Gmail and Docs, according to 9to5Google.
AI Pro subscribers lose Argon, and API users get a cheaper default after November 30
If you use Gemini through AI Pro, pick Neon for research, long documents and analysis, and keep Hydrogen for quick questions. Carbon still needs AI Ultra, which starts at $99.99 a month, and Google says Neon matches or beats it on coding, reasoning and agent tasks at the $19.99 tier. That makes the upgrade to Ultra harder to justify for a solo founder.
If your product runs on Argon through the API, test Neon now. Argon’s price doubles after November 30, and Neon will then cost 25% less per token. Run your own evaluations on real prompts before you switch, because every benchmark above comes from Google’s announcement.
If your team uses Gemini through Google Workspace, ask your admin when Neon reaches your plan. Google’s post names Gemini for business, but not which Workspace editions get the model menu.
