---
title: "Google’s Gemini 4 Neon costs $3 per million input tokens and ranks third on an AI intelligence index"
url: https://badiyaa.com/gemini-4-neon-google-model-price/
published: 2026-11-19T04:16:00+00:00
modified: 2026-11-19T04:16:00+00:00
section: Tech
publisher: Badiyaa
---

# Google’s Gemini 4 Neon costs $3 per million input tokens and ranks third on an AI intelligence index

Google's Gemini 4 Neon launches at $3 per million input tokens with a one-million-token context window, sitting between Carbon and Hydrogen.

Google has slotted a fourth model into its Gemini 4 family in barely a month. Gemini 4 Neon, announced on November 18 by Google DeepMind’s Koray Kavukcuoglu, is pitched as the middle child: quicker and cheaper than the flagship Carbon, deeper than the speedy Hydrogen, and priced at $3 per million input tokens and $15 per million output tokens.

Google says Neon scored 61 on the Artificial Analysis Intelligence Index at its highest thinking setting, which the company puts at third place overall, behind bigger and slower systems. It ships with a context window of one million tokens and is live now for developers and paying subscribers.

## What Gemini 4 Neon is for

Google frames the Gemini 4 lineup by job. In the company’s own description, Carbon handles the hardest long-horizon projects, Hydrogen covers fast everyday work, and Neon sits between the two for coding, analysis and enterprise workflows that run often and need solid reasoning without flagship prices.

The launch post leans hard on agent work. Neon is meant to run long, tool-heavy jobs from start to finish, either alone or as an “Orchestrator” that hands pieces of a task to parallel Hydrogen subagents inside Google Antigravity, the company’s agent-first coding platform. Google says it reads fewer files it does not need, reports progress as it goes, and copes with context compaction in long sessions.

## The numbers Google is quoting

The benchmark table on the Google DeepMind Gemini page compares Neon against two rival flagships. A few figures stand out:

- 70.9% on the Vals Index, which Google ranks second.

- 65.2% on Terminal-Bench 4.1 and 46.2% on FrontierCode 2.0 for coding.

- 39.0% on SWE-sweep when run in Antigravity, against 12.8% on the simpler mini-SWE harness.

- 84.0% on RiemannBench and 71.3% on HLE-Diamond for math and reasoning.

- 95.2% on GraphWalks, a long-context test spanning 256,000 to one million tokens.

These are the company’s own reported results, and several of the benchmarks are new or run on Google’s own tooling, so outside testing over the coming weeks will matter more than launch-day tables.

## How the price compares inside the family

A quick calculation shows what the rates mean in practice. A job that feeds Neon 200,000 tokens of documents and gets back 10,000 tokens of output would cost about $0.60 for input plus $0.15 for output, or roughly $0.75 at list price. Cached input drops to $0.15 per million tokens, a 95% discount, so repeated runs over the same files get far cheaper. For comparison, our earlier piece on [Google’s Gemini 4 Hydrogen launch at $1.50](https://badiyaa.com/gemini-4-hydrogen-google-launch/) covered the faster, lighter sibling.

Google also claims that at its default “high” setting Neon costs 32% less per task than Carbon while keeping a strong intelligence score, and that its top “xhigh” mode still costs at least 36% less per task than rival models at their highest settings. Neon has four thinking levels in total, letting teams trade depth for token spend.

## Security and safety claims

A separate “Neon for cyber” variant targets defenders. Google reports state-of-the-art scores of 90.5% on SEC-bench Pro for finding vulnerabilities and 51.26% on WebExploitBench v1.2, plus 70.7% on CWE-bench v2 for patching. That version reaches vetted security teams through the Fairwind Program over the next few days.

On behavior, Google says Neon took shortcuts on 2.45% of tasks in its coding cheating evaluation, the lowest of any Gemini 4 model, and passed 99.2% of its agent honesty tests.

## Who can use it

Neon is rolling out in the paid tier of the Gemini API through Google AI Studio and Android Studio, in Gemini for business, in Antigravity for paid Google AI Pro and Ultra subscribers, and in the Gemini app model menu for Pro and Ultra users. If you use the app, our guide to [Gemini’s Temporary Chat and Keep Activity settings](https://badiyaa.com/google-gemini-temporary-chat-keep-activity/) explains how to keep conversations out of training, and Sergey Brin’s comments on Google’s [“RSI treadmill” of Gemini models helping build their successors](https://badiyaa.com/sergey-brin-gemini-rsi-treadmill-argon/) explain why releases are arriving this fast.

Neon follows Argon in September, Carbon in October and Hydrogen earlier this month, and Google says the Fairwind cyber version starts reaching approved defenders within days.

*Image: Chetvorno, CC0 via Wikimedia Commons*

*This article was produced with AI assistance and reviewed by our Editorial Desk. [How we work](https://badiyaa.com/editorial-policy/).*
