Skip to content

Gemini 4 Neon is priced at $3 per million input tokens, and the Gemini app needs Pro or Ultra

Google's Gemini 4 Neon costs $3 per million input tokens and $15 per million output tokens, and Gemini app access is limited to Pro and Ultra subscribers.

Gemini 4 Neon: $3 Pricing, Availability and Benchmarks

Google began rolling out Gemini 4 Neon on November 18, 2026, pricing the model at $3 per million input tokens and $15 per million output tokens in the Gemini API. In the Gemini app, Pro and Ultra subscribers can pick it from the model dropdown.

Neon sits between the heavier Carbon and the faster Hydrogen in the Gemini 4 family, and Google is selling it on cost per task as much as on raw scores. For developers, that turns the buying question from which model is smartest into which one finishes a job for the least money.

Where Gemini 4 Neon is available

Google’s announcement, published at 4:00 p.m. ET on November 18, names the Gemini API paid tier in Google AI Studio and Android Studio as the main entry points. Gemini for business gets it through the API or its model dropdown, and Google Antigravity carries it too.

Consumers reach Neon through the Gemini app, where Pro and Ultra subscribers can select it. The post does not mention free Gemini app users, so the model picker looks limited to paid plans for now.

Eligible members of Google’s Fairwind Program, which is aimed at cyber defenders, get access over the next few days. Developers paying per token will care more about the next part.

Gemini 4 Neon pricing and context window

Google lists $3 per million input tokens, $0.15 per million cached input tokens and $15 per million output tokens. Cached prompts get a 95% discount, which helps agents that resend the same long instructions or codebase on every step.

The context window is 1 million tokens. According to the model card, Neon accepts text, images, audio and video files, and its only output is text. Default synchronous output tops out at 64,000 tokens, with up to 1 million through background long decode, and the knowledge cutoff is March 2026.

The model has four thinking levels. High is the default, and xhigh adds deeper multi-step reasoning for the hardest engineering, analytical and scientific work. A team that leaves everything on xhigh will see a different invoice than one that matches the level to the job.

What the Gemini 4 Neon benchmarks claim

Every figure here comes from Google. Neon ranks third on the Artificial Analysis Intelligence Index with a score of 61 at xhigh, and second on the Vals Index at 70.9%. It scores 65.2% on Terminal-Bench 4.1 and 46.2% on FrontierCode 2.0 at the default high setting.

The cost claims carry the pitch. At high thinking, Google puts Neon at 32% lower cost per task than Carbon, and roughly 50% to 80% lower than top competing frontier models on the Vals Index and Agent Arena. At xhigh, it claims 36% or more savings per task against competitors at their highest settings. Those comparisons use Google’s chosen rivals and benchmarks, and independent runs will show whether they hold.

A model that needs fewer retries can undercut a cheaper per-token rate, and that is the argument Google is making with the per-task numbers.

On safety, the model card compares Neon with Gemini 4 Hydrogen and reports unjustified refusals down 3.00 percentage points, with other differences inside confidence intervals. Google lists the full Safety & Security Report as coming soon.

How Neon fits Google’s paid AI lineup

Neon joins Genie 4 in Project Genie as another Google AI launch sitting behind a paid plan. Gemini tasks went the other way and dropped its paywall in six countries.

For people who use Gemini mostly for everyday questions, the dropdown change is optional. The pricing and the cost per task figures are aimed at teams running coding and research agents all day.

Leave a Reply

Your email address will not be published. Required fields are marked *