Google's Gemini 4 Neon: dearer than Carbon today, 25% cheaper from 1 Dec
Google released Gemini 4 Neon on 18 Nov in Gemini for business and to Pro plans. It costs 50% more than Carbon now and 25% less after Carbon's 1 Dec rise.
Google released Gemini 4 Neon on 18 November 2026, a model it places between its largest, Gemini 4 Carbon, and its cheapest, Gemini 4 Hydrogen. Google's launch post says it is rolling out the same day in Gemini for business "via the API or model drop-down", on the Gemini API's paid tier, and in the Gemini app for Google AI Pro and Ultra subscribers. It follows Carbon on 10 November and Hydrogen on 12 November.
For an IT lead who administers Gemini for business, that is one more choice in the model menu, with no seat-price change announced. For whoever pays usage bills on Carbon or Argon, directly or through a vendor, Neon is a second way round the 1 December price rise, pitched at harder work than Hydrogen.
What it costs against the rest of the family
Google prices Neon at $3 per million input tokens and $15 per million output tokens (tokens are the chunks of text a model reads and writes, and what usage-based AI bills count). Carbon and Argon are $2 and $10 until 30 November, then $4 and $20, as we reported on 10 November and The Decoder's price table repeats. Neon is therefore 50% dearer than them today ($3 ÷ $2) and 25% cheaper from 1 December ($3 ÷ $4 = 75% of the price). Hydrogen, at $1.50 and $7.50, is half Neon's rate. Cached input, text the model was sent before and is reusing, is $0.15 per million, a 95% discount, Google says.
The rate per token is not the bill. Neon has four "thinking levels", a setting for how long the model reasons before answering; more reasoning uses more tokens. Google says that at the default level, high, Neon cost 32% less per task than Carbon on the Artificial Analysis Intelligence Index, a third-party benchmark, and does not say which Carbon price that assumes. It gives no token count for a ticket, a contract or a spreadsheet.
What Google claims for office work
- Finance, tax and legal: Google says Neon ranks first on Vals AI's Tax Agent Bench and Legal Research Bench and second on the Vals Index at 70.9%, and first on most of Surge AI's finance, accounting and spreadsheet-modelling tests. It gives no scores for the tax and legal benchmarks. Hydrogen was first or joint first on both six days ago at 52.3% and 55.8%, so a person still checks the output.
- Made-up facts: Google reports its "lowest fabrication rate yet" in internal testing and gives no number.
- Hidden instructions: Google says that with its safeguards on and at its highest thinking level, Neon resisted every attack on Gray Swan's benchmark of indirect prompt injection (instructions planted in a web page, email or document to hijack the assistant). The figure is Google's.
- Plans: Carbon reached the Gemini app for Ultra subscribers only; Neon is on Pro as well, and Tasks in Gemini, the mode that works through multi-step chores, can run on it.
Reading this at work? Get one story like it explained every morning.
A five-minute read every morning at 7am ET: one story explained, two in brief, what else happened, one number. Unsubscribe in one click.
The catch
The post names no customer and gives no error rate on real office work. It does not mention Google Sheets, Docs or Search, does not say which model is the default in Gemini for business or whether an admin can set one, and is silent on training staff and on who reviews output.
On Hacker News, the commonest complaint is that four element names and four thinking levels do not say which model to use for what. One commenter adds that Neon scores about level with OpenAI's GPT-6.2 Sol on Artificial Analysis's index; Sol lists at $2 and $10, as we reported on 17 November, so Neon's rate is 50% higher. We have not checked that comparison.
What to do
- IT: list every job and vendor product billed at Carbon or Argon rates and ask what the bill becomes on 1 December. Re-run 50 to 100 real cases from each on Neon and on Hydrogen, and record the tokens used per finished task, not the rate.
- IT: tell Gemini for business users in one line which model to pick for which work, since Google has not said what the default is.
- Finance: before relying on Gemini for tax, close or spreadsheet work, ask Google for the Tax Agent Bench score behind the first-place claim.
Take this to work
Forward this
Google released Gemini 4 Neon on 18 November 2026 at $3 and $15 per million input and output tokens, 50% above Gemini 4 Carbon and Argon today and 25% below them once they double on 1 December, by Google's launch post. Any job we pay Carbon or Argon rates for, directly or through a vendor, should be re-tested on Neon and on the cheaper Hydrogen before that date.
Questions to ask
- Which of our vendors' products run on Gemini 4 Carbon or Argon, and will their price to us change on 1 December?
- How many tokens does one of our typical tasks use on Neon at the default thinking level, against Carbon and Hydrogen?
- Which model is the default in our Gemini for business menu, and can an admin set or lock it and the thinking level?
- What score did Neon get on Vals AI's Tax Agent Bench and Legal Research Bench, and who on our team reviews its output on that work?
- Does Neon reach Gemini in Sheets, Docs and Gmail on our plan, and when?
The arithmetic
(Input tokens per month ÷ 1,000,000 × input price) + (output tokens per month ÷ 1,000,000 × output price), worked at Neon's prices and at the prices of the model you use now, using the tokens each model needs to finish the same task.
From Google's post and our earlier reports: 1 million input and 200,000 output tokens cost $3 + $3 = $6 on Neon, $1.50 + $1.50 = $3 on Hydrogen, $2 + $2 = $4 on Carbon or Argon until 30 November and $4 + $4 = $8 from 1 December. The volumes are an illustration; Google gives no usage per task and models differ in tokens used, so take yours from a test run.
Questions this article answers
How much does Gemini 4 Neon cost?
Google prices it at $3 per million input tokens, $0.15 per million cached input tokens and $15 per million output tokens on the API, as of 18 November 2026. Google announced no change to Gemini or Workspace subscription prices.
Is Gemini 4 Neon cheaper than Gemini 4 Carbon?
Not until December. Carbon and Argon cost $2/$10 per million tokens until 30 November 2026, so Neon is 50% dearer; from 1 December they cost $4/$20 and Neon is 25% cheaper. Google also says Neon cost 32% less per task than Carbon on one third-party benchmark.
Who can use Gemini 4 Neon?
Google says it is rolling out from 18 November 2026 on the Gemini API paid tier, in Gemini for business through the API or the model menu, in Google Antigravity, and in the Gemini app for Google AI Pro and Ultra subscribers.
What is the difference between Gemini 4 Neon, Carbon and Hydrogen?
Google says Carbon is for the hardest long-running projects, Hydrogen for fast everyday work, and Neon sits between them for complex coding, analysis and business workflows. Hydrogen costs half Neon's rate per token.
Citi and PNC tell customers they are liable for what AI agents do
Five large US banks now hold customers responsible for AI tools they let in, hacks included. Finance should ban shared bank logins; legal should set a rule.
Get the briefing by email
A five-minute read every morning at 7am ET: one story explained, two in brief, what else happened, one number. Unsubscribe in one click.