Skip to content
Skip to content

Independent e-magazine

the OUTSPOKEN digest

Google Launches Gemini 4 Neon at $3 per Million Input Tokens, Between Carbon and Hydrogen

The third Gemini 4 model in eight days promises Carbon-level work at lower cost, with a 1 million token window, four thinking levels and a price rise for older models due on 30 November.

Outspoken Digest Technology Desk

Wednesday, November 18, 2026/2 min read

A white-tiled corridor between two rows of black server racks in a supercomputing facility, photographed on 15 February 2017, a file photograph illustrative of the data centres that run large AI models and not of Google's hardware or Gemini 4 Neon
Photo: Oleg Alexandrov via Wikimedia Commons (CC BY-SA 4.0)

Google has filled the middle of its newest model family. On Wednesday 18 November the company released Gemini 4 Neon, which sits between Gemini 4 Carbon, built for the hardest long-running projects, and Gemini 4 Hydrogen, built for fast everyday work. It is the third Gemini 4 model in eight days.

Google's own announcement says Neon brings frontier-level intelligence with fast end-to-end task completion. The practical news is the price, and what happens to the older models around it.

How much does Gemini 4 Neon cost?

Neon costs $3 per million input tokens and $15 per million output tokens, according to Google's launch post. Cached input is $0.15 per million tokens, a 95 percent discount. At its default high thinking setting, Google says Neon costs 32 percent less per task than Carbon.

9to5Google adds the calendar. Argon and Carbon are on a promotional $2 and $10 until 30 November, after which they rise to $4 and $20, which would make Neon 25 percent cheaper on input and output. The site notes that Neon's price is twice Hydrogen's.

What can it do, and what does Google claim?

Neon has a 1 million token context window and four thinking levels, from low to extra high. Google reports that at the highest setting it scores 61 on the Artificial Analysis Intelligence Index, third overall, and tops that firm's coding agent index at 58.6. It also reports first place on several agent and professional benchmarks, including a legal research test and a tax agent test.

These are the company's own figures, published on launch day, and independent testing will take longer than a day. Google also reports a 0.0 percent success rate for indirect prompt-injection attacks at the highest setting on one third-party test, a number worth watching as agents handle real work.

On science and engineering tests Google reports 70.5 percent on Terminal-Bench Science and 71.3 percent on HLE-Diamond, and it says Neon showed no cheating on Terminal-Bench 4.0 and a 0.09 percent misaligned action rate on its own agentic safety suite. All of that is self-reported for now.

Where can people use Gemini 4 Neon?

Developers get it through the paid Gemini API tier in Google AI Studio and Android Studio, and in Google Antigravity for paying AI Pro and Ultra subscribers. Businesses reach it through Gemini for business. Consumers need an AI Pro or Ultra subscription to find it in the Gemini app, where, per 9to5Google, it replaces Argon and Gemini 3.1 Pro is going away.

A cyber variant will reach members of Google's Fairwind Program, a group of trusted defenders, over the coming days. The next date to watch is 30 November, when the promotion ends and customers learn how many of them move to the cheaper middle model.

Published in The Outspoken Digest

Editorial desk

Outspoken Digest Technology Desk

Software, hardware, artificial intelligence and what they change for everyone else.

Newsletter

The Digest, in your inbox

One edition, sent when it is ready. No noise, and your address is never passed on.

We send a confirmation first. One click to leave, always.

Share this story

the OUTSPOKEN digest

Beyond boundaries. Independent stories on technology, culture, and the trends shaping how we live.