Thursday, November 19, 2026

News

Google ships Gemini 4 Neon at $3 per million input tokens, its fourth new model in four weeks

ModelsPatryk Raba
Google ships Gemini 4 Neon at $3 per million input tokens, its fourth new model in four weeks
Fot. Asoundd, Wikimedia Commons (CC BY-SA 4.0)

Google DeepMind has released Gemini 4 Neon, a model positioned between Carbon and Hydrogen, with a 1 million-token context window and API pricing of $3/$15 per million tokens. It is the fourth Gemini 4 model in four weeks.

Contents
  1. Pricing and specs
  2. Benchmark results
  3. Where the model is available
  4. What it means for teams

On November 18, 2026, Google DeepMind announced Gemini 4 Neon, the latest model in the Gemini 4 family. Neon sits between the slower but more powerful Carbon and the cheaper Hydrogen, and is meant to pair high quality with a lower cost per task.

The announcement was signed by Koray Kavukcuoglu, SVP of Google DeepMind and Chief AI Architect. Google describes Neon as a model offering frontier-level intelligence and fast end-to-end task completion across software engineering, knowledge work and science.

Pricing and specs

According to The Decoder, Neon costs $3 per million input tokens and $15 per million output tokens in the API. That is twice the price of Hydrogen ($1.50 and $7.50). Carbon and Argon have an introductory price of $2 and $10, which rises to $4 and $20 on December 1.

The model accepts text, images, video, audio and PDF files, with a context window of about a million tokens and output of up to a million tokens. It supports Computer Use, code execution, file search and grounding in Google Search and Maps results, among other features. It does not generate images or audio.

Benchmark results

Google says that at xhigh, Neon scores 61 on the Artificial Analysis Intelligence Index, good for third place overall, and 70.9% on the Vals Index, second place. On the Artificial Analysis Coding Agent Index the model ranks first with 58.6. It reaches 65.2% on Terminal-Bench 4.1 and 46.2% on FrontierCode 2.0 at the default high setting.

The company also cites first places on Legal Research Bench, Tax Agent Bench, Zapier AutomationBench (56.3%) and Vending-Bench 2. These are results presented by Google in its announcement, not a separate independent measurement.

According to Google, cost per task at high is 32% lower than with Carbon. The company also says that at high, Neon sits on the Pareto frontier on the Vals Index and Agent Arena, with cost per task roughly 50% to 80% lower than top competing models. At xhigh, it is said to cost more than 36% less than competitors at their highest settings.

Where the model is available

Neon is available in the paid tier of the Gemini API in Google AI Studio and Android Studio, and in Gemini for business. In Google Antigravity it is available to paid AI Pro and Ultra subscribers, and in the Gemini app to Pro and Ultra subscribers. Tasks in the Gemini app can now run on Neon.

In Antigravity, Orchestrator mode lets users pair Neon as the planner with Hydrogen as the executor in a single session. A Neon version for cyber defenders will reach eligible members of the Fairwind Program in the coming days. Access to the program can be requested through your Google Cloud account team.

What it means for teams

For companies paying for the API, Neon is for now more expensive per token than Carbon at its introductory price. From December 1, when Carbon rises to $4 and $20, Neon will sit between Hydrogen and Carbon on price. It is worth calculating the cost of a whole task rather than the per-token rate alone, since that is the basis of Google's comparisons. Some of the results come from the vendor, so they should be checked on your own data before migrating.

Share: