What Is Gemini 4 Neon? Price, Access and Carbon Compared
Thinking levels, what the 32% lower cost per task means, the model card comparison with Claude Opus 5.5 and the KVKK questions for companies.
In short: Gemini 4 Neon is the model Google made generally available on November 18, 2026, and it sits between the fast Hydrogen and the most capable model, Carbon, in the Gemini 4 family. It is built to finish complex software engineering, knowledge work and science workflows faster than Carbon. In the Gemini API it runs on the paid tier only, and its Standard price is $3 per million input tokens and $15 per million output tokens. In the Gemini app, Google AI Pro and Ultra subscribers can pick it.
The Gemini 4 family now has four models, each aimed at a different kind of work. Three things decide whether Neon fits: price, access and how it compares with Carbon. Its token price is higher than Carbon's introductory price, which runs through November 30, but Google says it finishes the same work faster and with fewer output tokens. There is no free tier, so even a first test needs a Cloud project with billing turned on.
What is Gemini 4 Neon, and where does it sit in the Gemini 4 family?
#Gemini 4 Neon is a model Google DeepMind adapted from Gemini 4 Carbon and announced on November 18, 2026. Google's Introducing Gemini 4 Neon announcement positions it like this: Carbon takes the hardest, long-horizon projects, Hydrogen handles fast everyday execution, and Neon balances depth and efficiency. According to the announcement, Neon delivers frontier-level reasoning for complex, high-frequency workflows faster and more cost efficiently than Carbon.
The model became generally available (GA) the same day. According to the Gemini API release notes, the stable model ID is gemini-4-neon. Google DeepMind's Gemini 4 Neon Model Card gives a knowledge cutoff of March 2026. The Choosing a model page in Google AI Studio recommends Neon for work where Hydrogen is not enough but you do not need Carbon's extra verification.
How much does Gemini 4 Neon cost?
#On the paid Standard tier, Gemini 4 Neon costs $3 per million input tokens and $15 per million output tokens, thinking tokens included. As of November 20, 2026, the Gemini Developer API pricing page in Google AI Studio lists these options:
| Option | Input | Output (including thinking) | Cached 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 |
The pricing page says plainly that "Gemini 4 Neon is not available on the Free tier." Neon supports implicit caching only, with no explicit caching. According to Choosing a model, Neon has no introductory price either. Carbon and Argon's Standard price is $2 per million input tokens and $10 per million output tokens through November 30, 2026, rising to $4 and $20 from December 1, 2026.
Who can access Gemini 4 Neon, and where?
#Access to Gemini 4 Neon opens through the paid Gemini API for developers and through a Google AI Pro or Ultra subscription for end users. Google's announcement and the Gemini Help Center's About Gemini models page list access like this:
| Channel | Who |
|---|---|
| Gemini API | Through Google AI Studio and Android Studio, on the paid tier. Google AI Pro and Ultra subscribers can try it in AI Studio. |
| Gemini app | Google AI Pro and Ultra subscribers, from the model menu. Neon is not in the rows for users with no plan or with Google AI Plus. |
| Google Antigravity | Paid Google AI Pro and Ultra subscribers and enterprises |
| Gemini for business | Through the API and the model menu |
| Neon for cyber | Eligible cyber defenders in the Fairwind Program, over the next few days |
Türkiye is on the list of available regions for the Gemini API. On Google Cloud, the Gemini 4 Neon page for Gemini Enterprise Agent Platform says the model is available to verified enterprise customers in select regions and lists global, us and eu as supported regions. RuneLab's Google Cloud Vertex AI and Gemini Enterprise Agent Platform course covers choosing models and building agents on Agent Platform.
What are Gemini 4 Neon's technical specs?
#Gemini 4 Neon takes text, images, video, audio and PDF, and produces text only. According to the Gemini 4 Neon model page in Google AI Studio, the model does not generate images or audio and does not support the Live API. Input, thinking and output tokens share one 1,048,576-token window. The default output limit per request is 65,536 tokens. According to the What's new in Gemini 4 Neon page, background execution or client-side continuation can extend output up to 1,048,576 tokens minus your input tokens. If tokens are new to you, start with our post what is a token.
What are Gemini 4 Neon's thinking levels?
#Neon has four thinking levels: low, medium, high, which is the default, and xhigh. The What's new page recommends starting with high for most complex tasks and notes that xhigh increases latency and token use. According to the same page, if you run another Gemini 4 model today, switching to gemini-4-neon requires no request schema changes.
How does Gemini 4 Neon compare with Carbon, Argon and Hydrogen?
#All four Gemini 4 models share the same 1,048,576-token context window, and they differ in thinking levels, speed and price. According to Google's pricing, Choosing a model and model pages, they compare like this:
| Neon | Carbon | Argon | Hydrogen | |
|---|---|---|---|---|
| Google's positioning | Complex, high-frequency work, between Hydrogen and Carbon | Strongest reasoning model, hardest work | Efficient reasoning model with faster replies than Carbon | Default for everyday coding and agent work |
| Thinking levels | low, medium, high (default), xhigh | low, medium (default), high | low, medium (default), high | low, medium, high (default), xhigh |
| Standard input / output, USD per million tokens | 3 / 15 | 2 / 10 through November 30, 4 / 20 from December 1 | Same as Carbon | 1.50 / 7.50 |
| Free API tier | No | No | No | Yes |
Through November 30, Neon's token price is higher than Carbon's. The "32% lower cost per task than Carbon" in Google's announcement is per task, not per token, and it applies at Neon's default high thinking level. According to Choosing a model, Neon finishes faster and with fewer output tokens than Carbon. Once Carbon's price doubles on December 1, Neon will also be cheaper than Carbon per token.
Google's recommendations split the work like this. Neon is suggested for multi-file coding, dense document and spreadsheet analysis or science workflows where Hydrogen is not enough. Carbon is suggested for the hardest work, such as system architecture, isolating root causes across repositories and security audits. Hydrogen is enough for everyday coding, interactive apps, parallel subagent tasks and trying things on the free tier. For the details, see our posts what is Gemini 4 Hydrogen, our Gemini 4 Carbon guide on the code security side, and which Gemini 4 model to choose.
How does Gemini 4 Neon compare with Claude Opus 5.5?
#In the table on Google's model card, Neon leads Claude Opus 5.5 and GPT-6.2 Sol on some benchmarks and trails on others. The figures below are Google's own results. According to the announcement, Neon's score of 61 on the Artificial Analysis Intelligence Index is at its xhigh thinking level, and its FrontierCode score is at the default high level.
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| AA Index v5.0.1 | 61 | 59 | 57 |
| Vals Index v2.2 | 70.9% | 70.1% | 66.1% |
| Terminal-bench 4.1 (cheating-adjusted) | 65.2% | 65.8% | 61.5% |
| FrontierCode 2.0 | 46.2% | 47.3% | 43.8% |
| Agents' Last Exam v1.1 | 75.8% | 71.6% | 73.0% |
Because this table is one model provider's own measurement, test the models on your own workload before you choose. For OpenAI's new models, see our post what are GPT-6.2 Astra and Sol.
What is Gemini 4 Neon good at?
#According to Google, Gemini 4 Neon stands out most in coding, agentic workflows and cybersecurity. The announcement says Neon ranks first overall on the Artificial Analysis Coding Agent Index with 58.6 and second on the Vals Index. In Google Antigravity it can also act as a fast orchestrator that directs parallel Hydrogen subagents. If agents are new to you, start with our post what is an AI agent.
An orchestrator's job is to give hard decisions to a strong model and routine steps to a fast one. To build the same idea into your own applications, RuneLab's RuneGate offers one OpenAI-compatible endpoint that routes each request by difficulty, sending simple requests to an economical model and hard ones to a powerful one, logs each request's model and cost, and supports OpenAI and Google Gemini models. RuneLab's Building Applications with LLM APIs course covers connecting the Gemini API to your own application.
On the cybersecurity side, Google says Neon scores 90.5% on SEC-bench Pro. Neon for cyber, the cyber defense version, will open to eligible teams in the Fairwind Program over the next few days.
What matters for company data and KVKK?
#Because Neon runs on the paid tier only, the data rules of the paid services apply. The "Used to improve our products" row of the pricing table says "No" on the paid tier. Under the Gemini API Additional Terms of Service in Google AI Studio, Google does not use your prompts or responses to improve its products on the paid services. Prompts and responses are logged for a limited period, solely to detect and prevent violations of the Prohibited Use Policy and for any required legal or regulatory disclosures. Under the terms, this data may be stored transiently or cached in any country where Google or its agents maintain facilities.
The paid tier does not settle the KVKK question on its own. Your company has to make its own assessment, including transfers of personal data abroad. This section is not legal advice. For a framework, see our KVKK generative AI guide, and for the choice between a cloud and an in-house model, our post local LLM vs ChatGPT. According to the model card, Neon is served through an API in the Gemini app, the Gemini API, Google AI Studio, Gemini for business and Google Antigravity, and there is no version a company can install on its own servers.
Work where data must never leave the company needs a model that runs in-house. RuneBox is a ready on-premise LLM installation on your company network. Models, RAG and agent flows run on an NVIDIA GPU box in a closed network with no outside connection. Because data is processed inside the company boundary, your BDDK and KVKK assessments become easier. Compliance is your organization's own assessment.
Frequently asked questions
#What is Gemini 4 Neon?
#Gemini 4 Neon is the model Google made generally available on November 18, 2026, and it sits between Hydrogen and Carbon in the Gemini 4 family. It is adapted from Gemini 4 Carbon and built to finish complex, high-frequency work faster than Carbon.
Is Gemini 4 Neon free?
#No. The Gemini API pricing page says Neon is not available on the free tier, and in the Gemini app it needs a Google AI Pro or Ultra subscription. In the Gemini 4 family, Hydrogen is the model you can use on the free tier.
What is the Gemini 4 Neon API price?
#As of November 20, 2026, the paid Standard tier costs $3 per million input tokens, $15 per million output tokens including thinking, and $0.15 for cached input. Batch and Flex cost $1.50 for input and $7.50 for output, and Priority costs $5.40 for input and $27 for output. Neon has no introductory price.
How is Gemini 4 Neon different from Carbon?
#Carbon thinks longer for the hardest work, while Neon is adapted from Carbon and finishes faster with fewer output tokens. Google says that at its default high level, Neon's cost per task is 32% lower than Carbon's, but through November 30 its token price is higher than Carbon's.
Gemini 4 Neon or Claude Opus 5.5?
#On Google's model card, Neon and Claude Opus 5.5 score 61 and 59 on AA Index v5.0.1, 70.9% and 70.1% on Vals Index v2.2, 65.2% and 65.8% on Terminal-bench 4.1, and 46.2% and 47.3% on FrontierCode 2.0. These are Google's own results, so test both models on your own workload.
How large is Gemini 4 Neon's context window?
#1,048,576 tokens. Input, thinking and output tokens share this window, and the default output limit per request is 65,536 tokens.
Is Gemini 4 Neon in the Gemini app?
#Yes, Google AI Pro and Ultra subscribers can pick it from the model menu. In the Gemini Help Center's plan table, Neon is not in the rows for users with no plan or with Google AI Plus.
What should companies in Türkiye watch for under KVKK?
#Neon runs on the paid tier only, and on the paid tier prompts are not used to improve Google's products, but the logs may be stored transiently or cached in any country where Google maintains facilities. The paid tier does not settle the KVKK question on its own, and the company must make its own assessment, including transfers abroad. This answer is not legal advice.
Sources
#- Google, Introducing Gemini 4 Neon: announcement, positioning, access, Artificial Analysis and Vals Index rankings, 32% lower cost per task, the orchestrator, SEC-bench Pro and Neon for cyber
- Google AI Studio, Gemini Developer API pricing: Neon, Carbon, Argon and Hydrogen prices, the free tier and the "Used to improve our products" row
- Google AI Studio, Gemini 4 Neon model page: model ID, input and output types, token limits and caching
- Google AI Studio, What's new in Gemini 4 Neon: thinking levels, long output and switching
- Google AI Studio, Choosing a model: positioning, thinking levels, introductory prices and recommendations
- Google AI Studio, Gemini 4 Argon model page: Argon's positioning and thinking levels
- Google AI Studio, Gemini API release notes: general availability on November 18, 2026, and the model ID
- Google DeepMind, Gemini 4 Neon Model Card: adaptation from Carbon, knowledge cutoff, distribution channels and the results table
- Google Cloud, Gemini Enterprise Agent Platform, Gemini 4 Neon: enterprise access and supported regions
- Gemini Help Center, About Gemini models: models in the Gemini app by plan
- Google AI Studio, Gemini API Additional Terms of Service: data use, logging and storage on the paid services
- Google AI Studio, Gemini API available regions: Türkiye
