A mid-tier model that keeps pace with the flagship: Google releases Gemini 4 Neon for code and heavy analysis
Independent measurements put Neon third worldwide, and per task it is cheaper and faster than Google's own Carbon.
Google released Gemini 4 Neon on Wednesday, the third model in the Gemini 4 family after Carbon and Hydrogen last week. According to 9to5Google, Neon sits between the fast Hydrogen for everyday questions and the heaviest model, Carbon, and replaces Argon and the older 3.1 Pro. Google is aiming at programming and complex analysis and claims Neon matches or beats Carbon on code, reasoning and agent tasks, at a speed close to that of the former Flash models.
The model handles a context of up to about one million tokens, accepts text, images, audio, video and PDFs and offers four thinking levels, from low to xhigh, DataCamp writes. Developers pay 3 dollars per million input tokens and 15 dollars per million output tokens, double the price of Hydrogen. Carbon temporarily costs 2 and 10 dollars until 30 November and then 4 and 20 dollars, which makes Neon roughly a quarter cheaper from December. There is no free tier.
Independent firm Artificial Analysis gives Neon at its highest thinking setting 61 points on its intelligence index, one point more than Carbon, good for third place. A task costs an average of 6.44 dollars, about 17 percent less than with Carbon, and finishes around 40 percent faster. Neon does use more output tokens, so the price advantage comes from the rate. According to DataCamp, Google itself published no direct comparison with Carbon.
Google AI Pro and Ultra subscribers can use Neon in the Gemini app and in Antigravity, developers through the Gemini API, AI Studio and Google Cloud.