Google has released Gemini 4 Neon, a middle model in the Gemini 4 family that sits between Hydrogen and Carbon. It costs $3 per million input tokens and $15 per million output tokens, with $0.15 per million for cached input. Tokens are the chunks of text a model reads and writes, and they are how API usage is billed.
Neon has a 1 million-token context window and four effort levels. Google is rolling it out in the paid Gemini API through AI Studio and Android Studio, in Gemini for business, and in Antigravity for Pro and Ultra subscribers.
The price order changes on December 1
The family now has three models, and which one is cheapest depends on the date. According to Simon Willison’s write-up of the launch, current prices are Neon at $3 and $15, Carbon at $2 and $10, and Hydrogen at $1.50 and $7.50. Carbon is on a launch discount, so today Neon is the most expensive of the three.
Willison says Carbon doubles to $4 and $20 on December 1, when its introductory pricing expires. At those prices Neon would cost 25% less than Carbon for the same input and output. Hydrogen, which we covered a week ago, currently costs less than both.
Carbon users get a doubled bill, and Neon is the cheaper step down
If an app or coding agent already runs on Carbon, the same usage costs twice as much after December 1. In Willison’s informal test, which drew the same pelican-on-a-bicycle picture on all three models, Neon’s result looked better than Hydrogen’s and worse than Carbon’s. That is one image, not a benchmark, so run a few of your real tasks on both Carbon and Neon and compare results and cost per task.
Effort level also changes the bill. On Willison’s drawing prompt, Neon produced about 4,974 output tokens at low effort and 9,691 at high effort, roughly 7.5 cents against 14.6 cents. If Neon’s price is a concern, start with a lower effort setting.
End of article