Technology

Add as a preferred source on Google

Google Debuts Gemini 4 Neon With a Combined 1 Million-Token Window

The model supports text, images, audio and video. API prices start at $3 per million input tokens, with access varying by channel and subscription tier.

Updated 2 min read
Google's headquarters stands among landscaped grounds in Mountain View, California. / TokenPost.ai (mono)
Google's headquarters stands among landscaped grounds in Mountain View, California. / TokenPost.ai (mono)

Google announced Gemini 4 Neon on Nov. 18, adding a model for complex coding, enterprise knowledge work and scientific tasks to a lineup aimed at developers and businesses. Its context window combines input and output capacity, reaching up to 1 million tokens across text, images, audio and video, according to Google's model card.

Google set API prices at $3 per million input tokens, $0.15 per million cached input tokens and $15 per million output tokens. The company listed access through the Gemini API, Google AI Studio, Gemini for business, Google Antigravity and the Gemini app, with availability depending on the channel and subscription tier.

Google describes Neon as positioned between Gemini 4 Carbon, designed for demanding long-running projects, and Gemini 4 Hydrogen, aimed at faster everyday tasks. The model card says Neon is adapted from Carbon.

For standard synchronous responses, Neon can generate up to 64,000 output tokens, while background long decoding can produce up to 1 million, Google says. Its model card reports scores of 61 on the Artificial Analysis Intelligence Index, 70.9% on the Vals Index, 65.4% on CursorBench 4.1 and 95.2% on GraphWalks, which tests inputs ranging from 256,000 to 1 million tokens. These are evaluations published by Google.

The model card gives March 2026 as Neon's knowledge cutoff and says the model may hallucinate. Google said a full safety and security report would be available in the coming days.

Related articles

View all