Launching something? List it on Upfynd for free.

Submit your product
AI Tools

Gemini 4 Neon: $3 in, $15 out, and when it beats Carbon on cost per task

Gemini 4 Neon lists at $3 input and $15 output per million tokens, half that on Batch and Flex. Carbon is cheaper per token until November 30, then doubles. Here is the break-even and where you can use Neon.

Akash Mohapatra · · Updated · 7 min read · 55 views

Gemini 4 Neon costs $3.00 per million input tokens and $15.00 per million output tokens on the Gemini API, as of November 2026. That is more per token than Gemini 4 Carbon today and less from December 1, when Carbon's list price doubles. So whether Gemini 4 Neon is the cheaper model depends on the date, how many tokens it spends to finish a task, and whether you can wait for batch pricing.

This post covers the prices on every tier, the arithmetic for comparing cost per finished task, and the places you can actually use the model: the paid Gemini API, Google AI Studio on paid plans, and GitHub Copilot.

What is Gemini 4 Neon?

Gemini 4 Neon is a Google model, API id gemini-4-neon, that Google places between Hydrogen and Carbon in its Gemini 4 line. GitHub's announcement says it is designed for agentic coding, multistep tool use, and knowledge work, and relays Google's claim of quality "close to Gemini 4 Carbon at a lower token cost per task." Google publishes no benchmark scores for it on its launch page, so that claim is Google's, not an independent measurement.

The Gemini 4 Neon model page lists these specs, last updated November 2026:

  • Context window: 1,048,576 tokens, shared by input, thinking and output.
  • Output: 65,536 tokens per synchronous request, up to 1,048,576 with long decode.
  • Inputs: text, image, video, audio and PDF. Output: text only.
  • Supported: function calling, code execution, computer use, structured outputs, URL context, file search, and grounding with Google Search and Google Maps.
  • Not supported: image generation, audio generation, the Live API and explicit context caching.

Thinking runs at four levels, low, medium, high and xhigh, and Neon defaults to high. Carbon and Argon default to medium and stop at high. That default matters for cost, because thinking tokens are billed as output.

Gemini 4 Neon pricing on every tier

These are Google's list prices per million tokens on the Gemini API pricing page, checked November 20, 2026. Output prices include thinking tokens.

TierInputOutputCached input
Standard$3.00$15.00$0.15
Batch$1.50$7.50$0.075
Flex$1.50$7.50$0.075
Priority$5.40$27.00$0.27

Batch and Flex cut every rate in half against Standard. Priority costs 80% more. Caching is implicit only: Google says it applies by default to prompts of at least 4,096 tokens, and cached input is billed at a 95% discount.

Grounding is billed separately. Google Search grounding includes 5,000 free requests a month, shared across Gemini 3 and later models, then costs $14 per 1,000 requests, with a default cap of 1,500 requests per day per project.

Is Gemini 4 Neon free?

No. Google says Neon is available on the paid tier only, not the free tier, of the Gemini API. In Google AI Studio it is offered to Google AI Pro and Ultra subscribers. If you need a Gemini 4 model with a free API tier, the pricing page lists Gemini 4 Hydrogen as free of charge on the free tier, with the condition that content on that tier is used to improve Google's products.

Gemini 4 Neon vs Carbon: the price flips on December 1

Carbon is cheaper per token until the end of November. Google's pricing page lists Gemini 4 Carbon at $2.00 input and $10.00 output through November 30, 2026, rising to $4.00 and $20.00 from December 1. Batch, Flex and Priority rates for Carbon double on the same date. The same page shows no scheduled change for Neon.

Model, Standard tierInput per 1MOutput per 1MNeon's per-token price vs it
Gemini 4 Hydrogen$1.50$7.502x
Carbon to Nov 30$2.00$10.001.5x
Neon$3.00$15.00-
Carbon from Dec 1$4.00$20.000.75x

Because Neon's rates are exactly 1.5 times Carbon's current rates and 0.75 times Carbon's December rates, on both input and output, the break-even is simple:

  • Until November 30, Neon is cheaper per task only if it finishes using less than two thirds of the tokens Carbon would.
  • From December 1, Neon is cheaper per task unless it uses more than a third more tokens than Carbon on the same job.

The December 1 increase also hits Gemini 4 Argon. We covered the full list in our post on the Gemini API price increase.

What a finished task costs: one worked example

To make the numbers concrete, take a hypothetical agentic coding run that reads 200,000 input tokens and writes 20,000 output tokens, thinking included. These token counts are an illustration, not a measurement.

  • Neon, Standard: $0.60 input plus $0.30 output, $0.90 in total.
  • Neon, Batch or Flex: $0.45.
  • Carbon to November 30: $0.60.
  • Carbon from December 1: $1.20.

If 150,000 of those input tokens hit the implicit cache, Neon's Standard cost falls to about $0.47. Repeated system prompts and repository context are the usual candidates for that discount.

The catch is the thinking default. If Neon at high thinking writes noticeably more output than Carbon at medium on your tasks, the per-token saving shrinks. Measure tokens per finished task on a sample of your own work before switching, and test Neon at medium as well as high.

Gemini 4 Neon in GitHub Copilot

GitHub made Neon generally available in Copilot on November 18, 2026 for Copilot Pro+, Max, Business and Enterprise. The plain Pro plan is not on that list. On Business and Enterprise, administrators control access through the model policy in Copilot settings, and new models are enabled by default unless an admin has turned that off. GitHub says the rollout is gradual.

The model is billed at provider list pricing under usage-based billing. GitHub's models and pricing table shows Neon at $3.00 input, $0.15 cached input and $15.00 output per million tokens, the same as Google's Standard tier, with 1 AI credit equal to $0.01. In Copilot, the worked example above would cost about 90 credits. GitHub's table lists no batch rate for it.

The same table marks Carbon's $2.00 and $10.00 rates as promotional until December 1, 2026, after which Copilot also charges $4.00 and $20.00. So the December flip applies inside GitHub Copilot too.

GitHub lists the model picker in VS Code, Visual Studio, JetBrains IDEs, Xcode, Eclipse, Copilot CLI, the Copilot coding agent, the Copilot app, github.com and GitHub Mobile. For how Copilot access shows up on a company bill, see our earlier post on the GitHub Copilot app and enterprise billing.

Which Gemini 4 model to pick

Google's own guidance is a useful starting point. Its Neon launch page recommends Carbon for the hardest architecture, debugging, science and security problems, keeps Hydrogen as the default for everyday coding and agentic workflows, and suggests Gemini 3.5 Flash-Lite for high-volume, low-latency work.

  • Pick Neon for multi-step coding agents and long-context work from December 1, when it is cheaper per token than Carbon, or now if you can run the work through Batch or Flex.
  • Stay on Carbon until November 30 if your tasks are short and you need its quality today. It is the cheaper per-token option until then.
  • Skip Neon if you need image or audio output, the Live API, explicit context caching, or a free API tier. Hydrogen is half Neon's price per token if its quality is enough.

Moving an existing Carbon, Argon or Hydrogen integration over takes no request schema changes, according to Google: you swap the model id to gemini-4-neon and check the thinking level. Integrations from Gemini 3 or earlier need more work, including removing temperature, top_p and top_k and setting client timeouts of at least 15 minutes.

Run the numbers on your own tasks

Before November 30, log input, output and cached tokens per finished task for Carbon on a sample of real jobs, then rerun the same sample on Neon at medium and high thinking. Multiply by the December rates, not today's. If Neon lands under 1.33 times Carbon's token count, it is the cheaper model from December 1.

If you are comparing hosted model providers more broadly, Upfynd lists 480 products in LLMs & AI Infrastructure and 447 AI coding tools, as of November 20, 2026. If you build one, you can list it for free.

Frequently asked questions

How much does Gemini 4 Neon cost?

Gemini 4 Neon costs $3.00 per million input tokens and $15.00 per million output tokens on the Gemini API Standard tier, with output including thinking tokens. Batch and Flex halve those rates to $1.50 and $7.50, Priority charges $5.40 and $27.00, and implicitly cached input costs $0.15 per million on Standard, as of November 2026.

Is Gemini 4 Neon free?

No. Google offers Gemini 4 Neon only on the paid tier of the Gemini API, not the free tier, and in Google AI Studio to Google AI Pro and Ultra subscribers. In GitHub Copilot it is billed at Google's list price under usage-based billing. Gemini 4 Hydrogen does have a free API tier.

Which Copilot plans include Gemini 4 Neon?

GitHub made Gemini 4 Neon generally available on November 18, 2026, for Copilot Pro+, Max, Business and Enterprise. The plain Pro plan is not listed. Business and Enterprise admins control it through the Copilot model policy, usage is billed at $3 input and $15 output per million tokens, and rollout is gradual.

What is the difference between Gemini 4 Neon and Carbon?

Google recommends Gemini 4 Carbon for the hardest problems and places Gemini 4 Neon below it, with high thinking as Neon's default. Carbon costs $2 and $10 per million tokens until November 30, 2026, then $4 and $20. Neon stays at $3 and $15, so it becomes cheaper per token from December 1.

What is the Gemini 4 Neon context window size?

Gemini 4 Neon has a 1,048,576-token context window that input, thinking and output share. A synchronous request can return up to 65,536 output tokens, and long decode raises that to the full 1,048,576. The model accepts text, image, video, audio and PDF input and returns text only.

Products in this article

About the author

Akash Mohapatra profile photo

Akash Mohapatra

Director & Co-founder

Connect on LinkedIn