Gemini 4 Neon: What It Does, What It Costs and How to Start

Where Gemini 4 Neon fits beside Carbon, Hydrogen and Argon, who can use it, and how to compare it with your current model on a real coding task.

Published

Tools
  • GGemini 4 Neon
Gemini 4 Neon: What It Does, What It Costs and How to Start

Gemini 4 Neon is Google DeepMind’s model for difficult work that comes up often, like debugging across several files or reconciling a spreadsheet with a contract. If your team builds on Gemini, Claude or GPT, try Neon on a recurring job where someone still spends time correcting the answer. Keep your current default until you know whether Neon saves that person work.

What Gemini 4 Neon does

Neon can work across spreadsheets, PDFs, charts, audio and video in a single 1 million-token context window. Think of the window as desk space. A token is a unit the model uses to represent material, and the window sets how much it can consider together. In a billing dispute, that leaves room for the contract, the invoices and the usage spreadsheet. Ask for the clause and spreadsheet row behind each disputed charge so a colleague can check the answer.

The thinking level controls how much effort Neon puts into the problem on that desk. The Google AI Studio changelog lists four settings for the generally available API model gemini-4-neon: low, medium, high (the default) and xhigh. Start at high. If the answer misses an invoice you left out, supply it. If Neon has the invoice but repeatedly misses how an amendment changes the rate, that is a reason to try xhigh.

A tray holds source documents while a separate thinking dial points to high.
Context determines what fits on the desk. Thinking level controls the effort spent working through it.

For a concrete example, Google showed Neon directing a group of agents in Antigravity, its coding environment. They used Blender to build a 3D lab bench with flames for hydrogen and carbon and gas discharge tubes for argon and neon. If you lead a development team, ask which parts of your own work can run in parallel. On a refactor, different agents might trace the callers of a function while one coordinates the final change.

Read Google’s Gemini 4 Neon launch announcement.

Which Gemini model fits the job?

Google introduced Neon on November 18, 2026. Its launch post places Neon between Hydrogen’s everyday execution and Carbon’s hardest long-running projects. Argon has some overlap with Neon. Google describes Argon in terms of the difficulty of the task and singles out Neon for difficult tasks you repeat frequently. These roles come from Google’s public launch post, API changelog and app notes, checked November 19, 2026.

ModelJob Google describesExample assignment
Gemini 4 HydrogenFast everyday coding and agent tasksUpdate a familiar test fixture.
Gemini 4 NeonComplex coding and analysis that need a quick turnaround, repeatedlyTrace a regression across a service and its callers.
Gemini 4 CarbonThe hardest projects that unfold over many stepsPlan a migration across dependent services.
Gemini 4 ArgonComplex coding, agent, math and science tasksWork through a mathematical model and explain its assumptions.

Where you can use Neon

For developers, Google lists the paid Gemini API in Google AI Studio and Android Studio. Google AI Pro and Ultra subscribers can also try Neon in AI Studio and Google Antigravity. If your team works in the latter, check which Antigravity plan covers your setup.

For app users, Google’s launch post names Google AI Pro and Ultra subscribers and directs them to the Gemini app’s model dropdown. A free Tasks rollout can be easy to confuse with free Neon access. The November 17 entry in the Gemini app release notes names Hydrogen as the model behind that rollout. The Gemini plan guide covers the subscription choices.

For enterprise teams, Google lists Neon in Gemini for business through both the API and the model dropdown. A buyer evaluating chat access should check the dropdown in the intended workspace. A developer budgeting for an integration needs the API rate.

On the standard paid API tier, each million tokens costs $3 for input, $0.15 for cached input or $15 for output, depending on which kind you use. Google’s Neon pricing

How to start with Neon

  1. Choose your entry point

    Open Google AI Studio with Neon selected. The link prompts you to sign in if you are signed out. For an API integration, use the model ID gemini-4-neon on the paid tier. For a chat trial with Pro or Ultra, open the Gemini app and select Gemini 4 Neon in the model dropdown.

  2. Pick a ticket you already closed

    Use the version of the code from before the fix. Give Neon the original bug report and the same files and tools your current model gets. Keep the accepted patch out of the prompt and reserve the regression test for scoring.

  3. Score the change like a reviewer

    Run the regression test against each patch. Read the changed files for unrelated edits and record how long it takes to get each answer ready to merge. A fast response that needs a rewrite can still cost you the afternoon.

Two paths lead from the same ticket through passing tests to a final measure of review time.
Use the same starting point and count the work left after the tests pass.

When to switch

Move a recurring job to Neon when it meets your acceptance criteria and consistently reduces time to an approved result at a cost you can sustain. For routine edits that Hydrogen already gets right, keep Hydrogen. For a long migration well served by Carbon, wait until you can isolate a repeatable part of the project for comparison. If OpenAI is also on your shortlist, the GPT-6.2 guide explains its model choices and access.

Choosing which model to pay for across your engineering team? Compare AI models for coding.

Published
Category
AI
Related Articles
Genie 4 Explained: Google's World Model and Who Can Use It

Genie 4 Explained: Google's World Model and Who Can Use It

Who can try Genie 4 in Project Genie, how live editing and audio work, and when creators and developers should use it for a prototype.Nov 19, 2026AI
Always-On Agents: Anthropic's Five Primitives, Shown Working

Always-On Agents: Anthropic's Five Primitives, Shown Working

An always-on agent mostly sleeps. Anthropic's five primitives wake it, carry its memory, guard its actions and deliver its work. Costs and limits included.Nov 19, 2026AI
Claude Cowork Pricing: What One Seat or a Team Costs

Claude Cowork Pricing: What One Seat or a Team Costs

Which Claude plans include Cowork? Compare US prices, usage limits, and worked budgets for one person, five seats, or twenty seats.Nov 19, 2026AI
QuillBot Review

QuillBot Review

When does QuillBot earn a seat beside your AI assistant? Compare its editing tools, US prices, team controls and privacy terms for client work.Nov 19, 2026AI
GPT-6.2 Astra vs Sol

GPT-6.2 Astra vs Sol

Choose a GPT-6.2 default by job. Compare Astra and Sol for agent runs, code review and writing, with worked examples and Ultrafast access by plan.Nov 18, 2026AI
GPT-6.2 Explained: Astra, Sol and Who Gets Access

GPT-6.2 Explained: Astra, Sol and Who Gets Access

Find GPT-6.2 Astra and Sol in ChatGPT Work, Codex and the API. See who gets Ultrafast and which jobs justify switching models.Nov 18, 2026AI
Kimi vs Qwen

Kimi vs Qwen

Kimi K3.1 Pro vs Qwen4-Max for coding agents and products: current access, long context, open weights, API tradeoffs and US deployment choices.Nov 17, 2026AI
Qwen 4: Qwen4-Max Explained

Qwen 4: Qwen4-Max Explained

Qwen4-Max explained: production versus preview access, context and output limits, open-model promises, and when to move from Qwen3.8-Max.Nov 17, 2026AI
Newsletter

One letter, every Sunday.Working systems, not hot takes.

Weekly. No spam. Unsubscribe anytime.