> ## Content Index
> Fetch the complete content index at: https://www.doitwithfriday.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Gemini 4 Neon Pricing: What One AI Automation Run Will Cost
- URL: https://www.doitwithfriday.com/news/gemini-4-neon/
- Published: 2026-11-19T09:25:00.000Z
- Updated: 2026-11-20T14:49:43.000Z
- Description: Google's new middle model comes with a published price. We turn it into the cost of one automation run, explain cached input and list what to ask your builder.
- Author: Kevin Anderson
- Tags: News, What It Costs

Google introduced Gemini 4 Neon on November 18, 2026, and the Gemini API release notes list it as generally available on the paid tier that day, at $3 and $15.

In Friday's view, an owner paying for AI automations should weigh the price of one run above a benchmark rank, and check that a workflow can move models later.

## Google releases Gemini 4 Neon on November 18

The announcement came from Koray Kavukcuoglu, SVP at Google DeepMind and Chief AI Architect at Google. The release notes list the model as generally available the same day, with pricing in the docs.

### What Google announced and where it runs

[Google's announcement](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/?ref=doitwithfriday.com) describes Gemini 4 Neon as bringing "frontier-level intelligence with fast end-to-end task completion" across software engineering, knowledge work and science. The model ID is gemini-4-neon. It is rolling out in the Gemini API through Google AI Studio, in Gemini for business through the API and the model drop-down, in Google Antigravity for paid Google AI Pro and Ultra subscribers, and in the Gemini app for Pro and Ultra subscribers. Google's post names Gemini for business and gives no list of Google Workspace plans that include the model, so check your own admin console before assuming your plan has it. The API pricing page says Gemini 4 Neon is not available on the free tier.

### Where Neon sits between Hydrogen and Carbon

Google's announcement places Neon against two Gemini 4 siblings. In its words, "Carbon tackles your hardest, long-horizon projects and Hydrogen handles fast everyday execution," while Neon balances depth and efficiency. Hydrogen reached GA on November 12 at $1.50 and $7.50 per million input and output tokens. Carbon launched on November 10 at an introductory $2 and $10\. Gemini 4 Neon sits between them in Google's lineup, yet for the rest of November it costs more per token than Carbon, a detail covered in the cost section below.

> GGoogle AI@GoogleAI
> 
> Meet Gemini 4 Neon — our newest model that brings frontier-level intelligence with fast end-to-end task completion across complex software engineering, knowledge work, and science workflows.
> 
> Neon brings frontier-level performance to your everyday work and is built to handle…
> 
> Nov 18, 2026[View on X](https://x.com/GoogleAI/status/2123168242797809944?ref=doitwithfriday.com)

Google AI's launch post on November 18, 2026, the same day the model went live on the Gemini API's paid tier.

### Benchmark claims are Google's own report

Google reports that Neon ranks third overall on the Artificial Analysis Intelligence Index with a score of 61 at its xhigh thinking level, and first overall on Zapier's AutomationBench at 56.3% and on Andon Labs' Vending-Bench 2, a simulation of running a business. Google also says Neon has a 32% lower cost per task than Carbon at the default high level. Every one of these figures comes from Google's announcement. In Friday's view, a public benchmark rank does not tell you how a model will do on your invoices or your support inbox, so treat these numbers as a reason to test, and test on your own work.

## Gemini 4 Neon pricing, line by line

The Gemini API pricing page lists Gemini 4 Neon in US dollars per million tokens, with four separate rates and one note about thinking tokens that decides how large the output side of a bill becomes.

### Input, output and thinking tokens

On the standard paid tier, Gemini 4 Neon costs $3.00 per million input tokens and $15.00 per million output tokens, with a 1 million-token context window, exactly 1,048,576 tokens shared by input, thinking and output. The pricing table labels the output line "including thinking tokens," and the page adds that every token the model processes is billed. Those tokens are the reasoning the model does before it answers. The pricing table charges them at the $15 output rate. Google's migration guide says Neon uses flat pricing across the full window, so a long prompt has no higher per-token rate.

### How cached input tokens get the 95% discount

Google's blog prices cached input tokens at $0.15 per million, "a 95% discount" on the $3 input rate. The catch sits in the pricing table: "Implicit caching only (explicit caching is not supported)." You cannot reserve a cache. The API spots a repeated prompt prefix on its own and discounts the cached input tokens when a request hits it. Google's caching guide says the minimum for Gemini 4 models is 4,096 tokens, and it gives two tips: put large and common content at the beginning of your prompt, and send requests with a similar prefix within a short time. The response reports cache hits in usage.total\_cached\_tokens, which is the number to check on your first invoice.

### Batch, Flex and Priority rates

The standard rate is one of four. Batch and Flex both cost $1.50 per million input tokens and $7.50 per million output tokens, half the standard price, and both list cached input at $0.075\. Priority costs $5.40 and $27.00, with cached input at $0.27\. Grounding with Google Search includes 5,000 free requests per month shared across Gemini 3 and newer models, then costs $14 per 1,000 requests. For an overnight job that nobody waits on, such as a weekly report, the half-price tiers are worth asking about before you pay standard rates.

## The cost per run of one automation, worked out

A per-token price only becomes useful once you multiply it by the tokens one run uses. Here is that arithmetic for one example workflow, using the standard paid-tier prices Google published.

### For example: an invoice-check run

For example, a bookkeeping workflow reads one supplier invoice, compares it with the purchase order and writes a short note. Each run sends 20,000 input tokens. Of those, 16,000 are the same instructions and reference rules every time, so they can bill as cached input tokens, and 4,000 are the new invoice. The run returns 2,000 output tokens, thinking tokens included. On Gemini 4 Neon, the cost per run is the sum of three lines, shown below with prices per million tokens from Google's pricing page.

```
Uncached input   4,000 x $3.00  / 1,000,000 = $0.0120
Cached input    16,000 x $0.15  / 1,000,000 = $0.0024
Output           2,000 x $15.00 / 1,000,000 = $0.0300
Cost per run                                 = $0.0444
1,000 runs a month                           = $44.40
Source: Gemini Developer API pricing, Gemini 4 Neon, standard paid tier
```

If the cache never hits, there are no cached input tokens, all 20,000 bill at $3, and the cost per run rises to $0.09, about double. Two choices separate those numbers: putting the fixed instructions first, and keeping them identical from run to run.

### The same run on Hydrogen and Carbon

Running the same tokens through the other Gemini 4 prices gives a cost per run of $0.0222 on Hydrogen and $0.0296 on Carbon at its introductory rate. Carbon's introductory rate runs through November 30\. From December 1, the pricing page lists Carbon at $4 input, $0.20 cached and $20 output, so the same run costs $0.0592\. Per 1,000 runs, that is $22.20 on Hydrogen, $29.60 on Carbon this month, $44.40 on Gemini 4 Neon and $59.20 on Carbon from December. Gemini 4 Argon, Google's efficient reasoning model, follows the same schedule: $2.00 and $10.00 through November 30, then $4.00 and $20.00\. A workflow built on Carbon or Argon in November doubles its token bill on a date its owner did not choose.

![Horizontal bar chart of the cost of 1,000 runs of one example automation at Gemini API standard paid-tier prices: Gemini 4 Hydrogen $22.20, Gemini 4 Carbon through November 30, 2026 $29.60, Gemini 4 Neon $44.40, and, highlighted, Gemini 4 Carbon from December 1, 2026 $59.20.](https://storage.ghost.io/c/67/23/6723d566-f4a4-4c03-8b45-4f8f0948342b/content/images/2026/11/gemini-4-neon-cost-per-run.webp)

For example, one run of 20,000 input tokens, 16,000 of them cached, and 2,000 output tokens costs $44.40 per 1,000 runs on Gemini 4 Neon, and Carbon's price doubles on December 1\. Standard paid-tier prices; Google's cost-per-task claims are not part of the chart. Source: Gemini Developer API pricing page, checked Nov 19, 2026.

### When the thinking level changes the bill

Gemini 4 Neon has four thinking levels: low, medium, high by default and xhigh. Google's guide says using xhigh increases latency and token use, and it advises starting with high and reserving xhigh for hard, multi-step problems. Because thinking tokens bill at the output rate, a run that thinks for another 4,000 tokens adds $0.06, more than the whole example run costs. Ask for the thinking level in writing, the reason it was picked, and the thinking tokens one run uses.

## What switching to Gemini 4 Neon takes

Google's migration guide splits the work in two. Moving from another Gemini 4 model takes a new model ID and two checks, while moving from Gemini 3 or older brings a longer checklist.

### From another Gemini 4 model, one model ID and two checks

Google says that if you run another Gemini 4 model today, switching to gemini-4-neon "requires no request schema changes." You set the model to gemini-4-neon, use a paid-tier project, and check the thinking level: Neon defaults to high and supports xhigh, while Carbon and Argon default to medium and cap at high. A workflow moved from Carbon without setting the level runs at a higher default level, and the extra thinking tokens show up in the output line of the bill. The change itself is small, in the request format of Google's examples:

```
"model": "gemini-4-neon",
"generation_config": { "thinking_level": "high" }
Source: Google, "What's new in Gemini 4 Neon," Gemini API docs
```

### From Gemini 3 and older, a checklist

[Google's migration checklist](https://aistudio.google.com/docs/whats-new-gemini-4-neon?ref=doitwithfriday.com) for Gemini 3 and earlier asks for more. You remove the temperature, top\_p and top\_k settings, declare tools through function calling instead of prompting for raw JSON, account for one shared 1M window for input, thinking and output, and raise client timeouts to at least 15 minutes, because hard prompts can take several minutes. Each item is a change to the code around the model, and each one needs a test run before the workflow touches real invoices or customer email.

### Explicit caching code has to go

If a workflow on Gemini 3 or 2.5 created its own caches, that code has to change before it moves to Neon. Google says Gemini 4 Neon does not support explicit caching and tells developers to remove cached\_content fields from requests, and its caching guide says no Gemini 4 model supports explicit caching. Implicit caching then takes over on prompts of at least 4,096 tokens, and the cached input tokens it finds bill at $0.15 per million. In Friday's view, this is where a cheap migration turns into a billable project, so ask before you agree to one.

## What to settle with the person who builds your automations

A new price on a new model is a good moment to check the deal you already have. Three questions cover lock-in, the token bill and the cost per run, and the answers belong in writing.

### Can you switch models without rebuilding?

Google's own guide shows the simple case: one model ID changes within the same family. That only helps you if the model name sits in one setting your builder can change, with the prompts and tests stored where you can see them. In Friday's view, an automation that hard-codes one vendor's model, caching calls and prompt format in several places is a lock-in cost you pay later, when prices move, as Carbon's will on December 1.

### Who pays the token bill?

A builder can set up the token bill in one of two ways. Either the API key and the Google Cloud billing account are in your name and you see every invoice, or the builder or platform resells tokens inside a monthly fee. Friday's recommendation is the first one: put the key in your name, set a budget alert, and ask for read access to usage, including usage.total\_cached\_tokens, so you can see whether the cache discount is real.

### Is the cost per run written down?

Ask for one sheet per workflow with new input tokens, cached input tokens, output tokens including thinking tokens, the thinking level, the model and the resulting cost per run, recalculated when a price changes. Our guide to [measuring AI automation ROI](https://www.doitwithfriday.com/blog/measure-ai-automation-roi/) shows how to set that cost against the time the workflow saves.

### When a small workflow beats a platform

In Friday's view, at $44.40 per 1,000 runs in the example above, the tokens are a small line next to the platform seat, the agency retainer and the rebuild when you change models. A business with two or three repeated tasks is better served by [a small workflow it owns](https://www.doitwithfriday.com/services/workflows/), with the model in one setting and the bill in its own name, than by a platform priced per seat.

## Before you choose a model for your next workflow

Gemini 4 Neon gives a published price that you can turn into a cost per run before you commit. Check the cache, the thinking level and the date Carbon's rate changes.

Write down the tokens for one real run, multiply them by the standard prices, and compare Hydrogen, Gemini 4 Neon and Carbon after December 1\. If you want a second opinion on a workflow's cost and portability, [send us one run you want priced](https://www.doitwithfriday.com/contact/).

Last checked: November 19, 2026, against Google's November 18, 2026 announcement of Gemini 4 Neon, the Gemini API release notes, the Gemini Developer API pricing page, Google's context caching guide and the "What's new in Gemini 4 Neon" migration guide.

Prices and model details come from Google's announcement and Gemini API documentation, which are the primary sources, checked November 19, 2026\. Benchmark figures are Google's own report. Friday has no relationship with Google. Statements marked as Friday's view or Friday's recommendation are opinion.