Google Gemini 4 Neon Launches: Price, Benchmarks and What It Means for You

Google has launched another new artificial intelligence model, and this one fills the gap its Gemini 4 family had left open. Gemini 4 Neon, announced on November 18, 2026, is built for the complex work teams run every day: multi-file coding, deep analysis of documents and spreadsheets, and long agent workflows that need to finish fast. It is not Google’s biggest model, and it is not its cheapest. It is the one Google expects most developers and knowledge workers to actually use.

Neon is the fourth member of the Gemini 4 family. Carbon, the flagship, arrived on November 10 for the hardest, longest jobs. Hydrogen, the fast and cheap everyday model, followed on November 12. Neon sits between them: frontier-level intelligence at a speed and price aimed at high-frequency professional work. It is available now in the Gemini API, Google AI Studio, the Gemini app and Google Antigravity.

Close-up of a humanoid robot with glowing blue lights representing Google Gemini Decision 1
Blue-lit robot. Photo by Kindel Media on Pexels.

What Is Gemini 4 Neon?

Gemini 4 Neon is a frontier-level model for software engineering, knowledge work and science. Google describes it as bringing “frontier-level intelligence with fast end-to-end task completion” across complex workflows — work where a model must plan, use tools, check its own output and keep going for many steps without losing the thread.

Neon accepts text, images, video, audio and PDF files as input, and it has a one-million-token context window shared between input, thinking and output. It can hold a large codebase, a stack of financial reports or a long legal file in mind at once, and turn dense material into structured deliverables: reports, summaries, spreadsheets and code.

Neon offers four thinking levels — low, medium, high and xhigh — so users can trade intelligence against speed and token spend for each job. High is the default.

One of Neon’s most interesting roles is as an orchestrator. In Google Antigravity, the company’s agentic development environment, Neon can act as a fast, high-judgment coordinator that directs parallel Hydrogen subagents — planning the work, handing out the routine steps, and checking the results.

The Benchmarks: Where Neon Leads

Google’s headline number is Neon’s score of 61 on the Artificial Analysis Intelligence Index at its highest xhigh thinking level — third overall, behind only larger, slower models. On Google’s own comparison table, that puts Neon ahead of Anthropic’s Claude Opus 5.5 at 59 and OpenAI’s GPT-6.2 Sol at 57.

The coding results are where Neon looks strongest. It ranks first overall on the Artificial Analysis Coding Agent Index with a score of 58.6, and first on the Vals Index at 70.9 percent. Google also reports that Neon leads on AutomationBench at 56.3 percent, and ranks first on the Vals Legal Research Bench and Tax Agent Bench — the tests that matter most to lawyers, accountants and analysts who work with dense documents all day.

On agentic work, Neon ranks second on steerability in the Agent Arena Index, a live benchmark drawn from more than 4,200 real agent sessions — a measure of how cleanly a model adapts when a user corrects it mid-task.

The picture is not one-sided. Google’s own figures show Neon behind Claude Opus 5.5 on Terminal-Bench 4.1 and FrontierCode 2.0, and behind GPT-6.2 Sol on Chartography. Every number here comes from Google, measured on benchmarks Google chose, and independent testing has not yet confirmed them. Treat the rankings as a strong first signal, not a final verdict.

Dual computer monitors displaying green code in a dark room
Code monitors. Photo by Tima Miroshnichenko on Pexels.

The Price: $3 In, $15 Out — and the November 30 Twist

Gemini 4 Neon costs $3 per million input tokens and $15 per million output tokens, with cached input at $0.15 per million — a 95 percent discount for repeated context. There is no free tier.

That makes Neon exactly twice the price of Hydrogen, which costs $1.50 and $7.50 per million tokens. But the comparison with Carbon is stranger. Carbon and Argon are on promotional pricing of $2 input and $10 output until November 30, 2026 — which means that, until the end of the month, Google’s flagship model is actually cheaper per token than Neon.

The comparison flips on a fixed date. When the promotion ends on December 1, Carbon rises to $4 and $20 per million tokens, and Neon’s standard rates become 25 percent lower per token without moving at all. Google says that at its default high thinking level, Neon already costs 32 percent less per task than Carbon, and 50 to 80 percent less per task than the leading frontier models.

A model that finishes a job in one attempt costs less than a cheaper model that needs three. Teams should watch their real bills, not just the price list.

Where Neon Sits in the Gemini 4 Family

Carbon is the flagship. It is built for the hardest, longest jobs: multi-hour software engineering, deep research and cybersecurity defence. It is the strongest model Google has, and also the slowest.

Hydrogen is the fast everyday model. It is built for rapid execution across routine coding and agentic work, and at $1.50 and $7.50 per million tokens it is the cheapest way into the family. It is the model Google calls its daily driver.

Neon sits between them. It is built for work that is complex but frequent: multi-file coding, analytical reasoning, spreadsheet and document analysis, and long agent runs that a team expects to repeat many times a day. Google says its quality is comparable to Carbon on coding, reasoning and agentic tasks — at Flash speeds and a lower cost.

Argon, the earlier frontier model, is being replaced by Neon in the lineup.

Our report on Google’s flagship model, Gemini 4 Carbon, explains what Carbon does best and why its price doubles on December 1. And our report on Gemini 4 Hydrogen covers the fast, affordable model that Neon now works alongside as an orchestrator.

Where You Can Use Gemini 4 Neon

Gemini 4 Neon is rolling out now across Google’s products. Developers can reach it through the Gemini API in Google AI Studio and Android Studio on the paid tier, and through Gemini for business, where it appears both in the API and in the model dropdown.

Paid Google AI Pro and Ultra subscribers get Neon in Google Antigravity and in the Gemini app, where it appears under “Deeper analysis” in the model picker.

There is no free tier for Neon. That is the clearest sign of where Google places it: a professional product for paying developers, businesses and subscribers who run complex work every day.

Google is also releasing a cyber-tuned variant, Gemini 4 Neon for cyber, built for secure coding, vulnerability discovery and patching. Google says it produces more secure code than Carbon across easy, medium and hard challenges on AutoBaxBench. It is rolling out to trusted defenders as part of Google’s Fairwind Program.

Who Should Use Gemini 4 Neon?

If you are a software developer, Neon is worth trying now. It is built for the work you do every day: repository-level debugging, multi-file refactoring, issue resolution and long agent sessions with many tool calls.

If you are a researcher, analyst or knowledge worker, the document abilities are the draw. Neon can synthesise spreadsheets, PDFs, charts, audio and video into polished, cited reports, and its top rankings on the legal and tax agent benchmarks suggest it handles dense professional material well.

If you are a casual user who mostly needs quick answers and short writing, Neon is not for you. You will not see it on the free tier, and Hydrogen or the standard Gemini app will be faster and cheaper for everyday questions.

The Catches to Keep in Mind

Neon has real limits, and the first is the benchmark question. Every figure in this article comes from Google, and independent verification will take time.

The second is the price twist. Until November 30, Carbon is cheaper per token than Neon on its promotional rate. Teams that sign up for Neon today should check whether their workloads actually run cheaper on Neon per task, as Google claims, before committing.

The third is the family itself. Four models with element names, four thinking levels, promotional prices that change on December 1, and a model picker that keeps changing — Google’s lineup is genuinely hard to follow. Developers will need to test Neon against Carbon and Hydrogen on their own work to know which one deserves their budget.

What Happens Next

The next few weeks will decide Neon’s reputation. Developers are already putting it through real projects, and their results — not Google’s launch benchmarks — will determine whether it becomes the default model for everyday professional coding or just another launch-day headline.

Watch whether developers keep using Neon after the novelty fades, and how Anthropic and OpenAI respond — GPT-6.2 arrived only yesterday, and the next answer is probably already being prepared.

In eight days, Anthropic shipped Fable 5.2, Google shipped Carbon, Hydrogen and Neon, and OpenAI shipped GPT-6.2. The AI race is moving faster than ever, and the winners will be decided not in the laboratory, but in the daily work of developers and businesses. Neon is Google’s bet that the middle of the market — complex work, done fast, at a fair price — is where that race will be won.

Leave a Comment