What Is Gemini 4 Hydrogen? Price, Access and Everyday Use
Thinking levels, moving from Gemini 3.8 Flash, pairing it with Carbon, the jobs where Hydrogen falls short and the KVKK questions for companies.
In short: Gemini 4 Hydrogen is the Gemini 4 model Google made generally available on November 12, 2026, built for speed and efficiency in everyday coding and agentic work. It takes text, images, video, audio and PDF, produces text and uses a single 1,048,576-token context window. It is available on the Gemini API free tier, and on the paid Standard tier it costs $1.50 per million input tokens and $7.50 per million output tokens.
Google splits the Gemini 4 family by job: Carbon and Argon for the hardest, long-horizon problems, and Hydrogen for fast execution. What sets Hydrogen apart from the rest of the family? Price and access come first: at $1.50 per million input tokens it sits below Carbon's and Argon's introductory price, and it is the only Gemini 4 model you can use on the free tier. It is a useful pick for everyday coding, tool-using agents and long documents, but not the first choice for hard architecture decisions or security audits. For those, Google recommends Carbon.
What is Gemini 4 Hydrogen, and when was it released?
#Gemini 4 Hydrogen is a model Google DeepMind adapted from Gemini 4 Carbon and announced on November 12, 2026. It became generally available (GA) the same day, and according to the Gemini API release notes its API model ID is gemini-4-hydrogen. Google's Introducing Gemini 4 Hydrogen announcement presents it as the model that brings Gemini 4 to everyone, built for speed and efficiency in everyday coding and agentic workflows.
According to the Gemini 4 Hydrogen model page in Google AI Studio, the model takes text, images, video, audio and PDF, and produces text only. Input, thinking and output tokens share a single 1,048,576-token window. The default output limit per request is 65,536 tokens, and with long decode the output can run into whatever the input leaves of the window, up to 1,048,576 tokens. Google DeepMind's Gemini 4 Hydrogen Model Card gives a knowledge cutoff of March 2026 and adds that in some domains the model's knowledge may be less up to date. If tokens are new to you, start with what is a token.
What are Gemini 4 Hydrogen's thinking levels?
#Hydrogen has four thinking levels: low, medium, high (the default) and a new xhigh. Google's What's new in Gemini 4 Hydrogen page recommends starting with high for most work and reserving xhigh for hard, multi-step problems where high falls short. The same page says xhigh raises latency and token use and can cause overthinking on simpler prompts. Carbon and Argon default to medium and top out at high, so xhigh exists only on Hydrogen.
Who can access Gemini 4 Hydrogen, and where?
#Gemini 4 Hydrogen opened to developers, enterprises and all users on the same day. In the Gemini API it is available on both the free and paid tiers, through GenerateContent and the Interactions API. Türkiye is on the list of regions where the Gemini API is available. Google's announcement lists access for three groups:
| Who | Where | Note |
|---|---|---|
| Developers | Gemini API, Google AI Studio, Android Studio | Paid Google AI Pro and Ultra subscribers can also try it in Google Antigravity |
| Enterprises | Gemini for business (API or model drop-down), Google Antigravity, Gemini in Google Sheets | Gemini in Sheets for eligible business subscriptions |
| Everyone | Gemini app model drop-down, Google Search | Gemini in Google Sheets, starting with Google AI Pro and Ultra subscribers |
Google Cloud's Gemini Enterprise Agent Platform model list also includes Hydrogen among its generally available Gemini models. RuneLab's Google Cloud Vertex AI and Gemini Enterprise Agent Platform course covers choosing models and building agents on Agent Platform.
Is Hydrogen in the free Gemini app?
#Yes. Users with no Google AI plan get Hydrogen in Tasks mode, and in chats Hydrogen steps in when auto model selection judges it the right model for the prompt. In the table on the Gemini Apps Help About Gemini models page, the row for users with no plan reads "Auto or Flash-Lite" for chats and "Hydrogen" for tasks. With auto selection on, most prompts use Flash-Lite and prompts that need more reasoning may use Hydrogen. With it off, every chat response uses Flash-Lite. Google AI Plus, Pro and Ultra list Hydrogen among both their chat and task models.
What is the Gemini 4 Hydrogen API price?
#On the paid Standard tier, Gemini 4 Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens including thinking tokens, flat across the full 1M-token window. As of November 18, 2026, the paid options on the Gemini Developer API pricing page in Google AI Studio are:
| Option | Input | Output (incl. thinking) | Cached input |
|---|---|---|---|
| Standard | 1.50 | 7.50 | 0.075 |
| Batch | 0.75 | 3.75 | 0.0375 |
| Flex | 0.75 | 3.75 | 0.0375 |
| Priority | 2.70 | 13.50 | 0.135 |
On the free tier, the Standard tab shows "Free of charge" for input, output and caching, and free use is rate-limited. Batch and Flex are not available on the free tier. Hydrogen supports implicit caching only, with no explicit caching. According to the What's new page, implicit caching applies by default to prompts of at least 4,096 tokens and cuts the price of cached input by 95% on a hit.
Grounding with Google Search is not available on the free tier. On the paid tier, 5,000 search requests a month are free, but that allowance is shared across all Gemini 3 and newer models. After that it costs $14 per 1,000 requests, with a daily cap of 1,500 requests. Grounding with Google Maps is not available on the free tier either.
The pricing page gives no end date for Hydrogen's price. Carbon's and Argon's Standard price, by contrast, is introductory and runs through November 30, 2026: $2 per million input tokens, $10 for output and $0.10 for cached input. From December 1, 2026, those prices rise to $4, $20 and $0.20. Neither model is offered on the free tier.
How is your data used on the free tier?
#On the free tier, the content you send is used to improve Google's products, but not on the standard paid tiers. The pricing table's "Used to improve our products" row says "Yes" for the free tier and "No" for the standard paid tiers (Standard, Batch, Flex and Priority). On the invite-only Contributor tier, the row says "Yes" on the paid side as well.
Under the Gemini API Additional Terms of Service in Google AI Studio, on unpaid services Google uses the content you submit and any generated responses "to provide, improve, and develop" its products, services and machine learning technologies. To help with quality, human reviewers may read, annotate and process your input and output. Google says it disconnects that data from your Google Account, API key and Cloud project before reviewers see it. The terms say, in bold, not to submit sensitive, confidential or personal information to the unpaid services.
The terms also define when use counts as paid. Gemini API use is a paid service only when requests go through a Cloud project associated with an active billing account. Google AI Studio counts as a paid service even when it is offered free of charge, as long as the account you use has access to a Cloud project with an active Cloud Billing account or is a Workspace enterprise account. For users in the European Economic Area, Switzerland and the United Kingdom, the paid services' data terms apply to all services, including Google AI Studio and the unpaid quota of the Gemini API. Türkiye is not on that list.
On paid services, Google does not use your prompts or responses to improve its products. It processes them under the Data Processing Addendum for products where Google is a data processor. 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. According to the terms, this data may be stored transiently or cached in any country where Google or its agents maintain facilities.
What is Gemini 4 Hydrogen good for?
#Gemini 4 Hydrogen is Google's default general-purpose model for everyday coding, interactive web apps, tool-using agents, multimodal document workflows and parallel subagent tasks. The What's new page calls it the recommended upgrade from Gemini 3.8 Flash and earlier Gemini 3 and 2.5 models. According to the announcement, Carbon and Argon excel at deep, long-horizon reasoning on the hardest problems, while Hydrogen is built for rapid execution across diverse agentic workflows. If agents are new to you, see what is an AI agent.
On the Artificial Analysis Intelligence Index, an independent benchmark, Hydrogen scores 50.8 at its default high level, above Argon's 50.4, and reaches 54 at xhigh. These scores appear on Artificial Analysis's Gemini 4 Hydrogen page, and Google's announcement gives the same comparison as 51 vs. 50.
What changes when you move from Gemini 3.8 Flash?
#By Google's own measurements, Hydrogen at its high default finishes tasks 2.5 times faster than Gemini 3.8 Flash and lifts Terminal-Bench 4.1 scores from 18.2% to 57%. Both figures come from Google's November 12 announcement. The model card labels the 57.0% Terminal-Bench 4.1 result "cheating-adjusted."
The move takes a few code changes. According to the migration checklist on the What's new page, the model ID changes from gemini-3.8-flash to gemini-4-hydrogen. The default thinking level becomes high instead of medium, and the minimal level returns an error. Explicit caching is not supported, so cached_content fields have to come out of requests. The temperature, top_p and top_k parameters should come out too: the server ignores in-range values, and out-of-range values return a 400 validation error. RuneLab's Building Applications with LLM APIs course covers connecting an LLM API such as the Gemini API to your own application.
How does Hydrogen work alongside Carbon?
#Google positions Hydrogen both as a standalone model and as the execution model next to Carbon. According to the announcement, Orchestrator, new in Google Antigravity, switches between models such as Carbon and Hydrogen within the same session. By Google's own measurement on Terminal-Bench 4.0, Orchestrator reached near-Carbon task success while using 24% less quota on average and finishing 19% faster. In a parallel agent setup, Carbon can act as the coordinator while Hydrogen runs the worker agents that write changes and the verifier agents that test them. To pick the model by task difficulty in 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, sets a monthly budget per team and supports OpenAI and Google Gemini models.
The Antigravity agent in Gemini Managed Agents now uses Hydrogen by default as well. We cover whether to give a job to one agent or to several working together in one agent or many.
When is Gemini 4 Hydrogen not enough?
#Hard architecture decisions, multi-file debugging and security audits are the jobs where Google recommends Carbon instead of Hydrogen. According to the Choosing a model page in Google AI Studio, Carbon spends more thinking tokens per turn to test hypotheses, isolate root causes across repositories and check fixes against tests. The page recommends Carbon for system architecture, multi-file debugging, dense multimodal analysis, formal proofs and security audits, or when a faster model gets stuck. For work where Hydrogen is not enough but you do not need Carbon's extra verification, the same page has pointed since November 18 to the model that sits between Hydrogen and Carbon, which we cover in what is Gemini 4 Neon.
The What's new page positions Argon for advanced long-horizon reasoning with lower token spend than Carbon. For high-volume classification, routing and sub-second latency targets, Google points to Gemini 3.5 Flash-Lite. The model card also acknowledges limitations such as hallucinations, occasional slowness or timeouts, and higher token use, especially at higher thinking levels. We cover Carbon's code-security side in our Gemini 4 Carbon guide and the choice among the three models in which Gemini 4 model to choose.
What matters for hosting and KVKK?
#Hydrogen is a cloud model and cannot be installed on a company's own servers. The model card describes it as a model offered through an API and distributed through the Gemini app, Gemini API, Google Sheets, Google AI Studio, Gemini for business, Google Antigravity and Google Search, and it offers no downloadable weights.
For a company using it from Türkiye, two rules stand out. On the free tier, content is used to improve Google's products, and the exception for the European Economic Area, Switzerland and the United Kingdom does not cover Türkiye. On the paid tier that use stops, but the logs may be stored transiently or cached in any country where Google maintains facilities. So do not run tests with personal data, customer records or confidential documents on the free tier. The terms ask the same.
Moving to the paid tier does not settle the KVKK question on its own. Your company has to make its own assessment, including any transfer of personal data abroad. This section is not legal advice. For the framework, see our KVKK generative AI guide. Work where data must never leave the company needs a model that runs in-house, and Hydrogen does not run that way.
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 Hydrogen?
#Gemini 4 Hydrogen is the Gemini 4 model Google made generally available on November 12, 2026, focused on speed for everyday coding and agentic work. It is adapted from Gemini 4 Carbon, takes text, images, video, audio and PDF, and uses a single 1,048,576-token window for input and output.
Can you use Gemini 4 Hydrogen for free?
#Yes, it is free of charge with rate limits on the Gemini API free tier and in Google AI Studio, but content sent on the free tier is used to improve Google's products. In the free Gemini app, Hydrogen is used in Tasks mode, and in chats it steps in when auto model selection picks it. Do not send personal or confidential data on the free tier.
What is the Gemini 4 Hydrogen API price?
#As of November 18, 2026, the paid Standard tier costs $1.50 per million input tokens, $7.50 per million output tokens including thinking, and $0.075 for cached input. Batch and Flex cost $0.75 for input and $3.75 for output, and Priority costs $2.70 for input and $13.50 for output. The pricing page gives no end date for Hydrogen's price.
How is Gemini 4 Hydrogen different from Argon and Carbon?
#Google positions Carbon and Argon for hard, long-horizon reasoning and Hydrogen for fast execution and everyday work. Hydrogen is also on the free tier and costs less, while Argon and Carbon are not on the free tier. Hydrogen thinks at high by default and offers an xhigh level that only it has.
What are Gemini 4 Hydrogen's thinking levels?
#Hydrogen has four thinking levels, low, medium, high and xhigh, and the default is high. Google recommends starting with high for most work and reserving xhigh for hard, multi-step problems where high falls short. The minimal level returns an error on Hydrogen.
What changes when you move from Gemini 3.8 Flash to Hydrogen?
#The model ID becomes gemini-4-hydrogen, the default thinking level becomes high instead of medium, and explicit caching has to be removed. In-range values of temperature, top_p and top_k are ignored, and out-of-range values return a 400 validation error. By Google's own measurement, Hydrogen finishes tasks 2.5 times faster than 3.8 Flash.
Can you install Gemini 4 Hydrogen on your own servers?
#No. Hydrogen is a cloud model, used through the Gemini API and Google's products, and it has no downloadable weights. Work where data must not leave the company needs a model that runs in-house.
What should companies in Türkiye watch for under KVKK?
#On the free tier, content is used to improve Google's products and human reviewers may read it, so do not send personal data. The paid tier drops that use, but the logs may be stored transiently or cached in any country where Google maintains facilities, and moving to the paid tier does not settle the KVKK question on its own. Your company must make its own assessment, including transfers abroad, and this answer is not legal advice.
Sources
#- Google, Introducing Gemini 4 Hydrogen: the announcement, positioning, access groups, the free Gemini app footnote, Google's comparison with Gemini 3.8 Flash, the Orchestrator measurement and the price
- Google AI Studio, Gemini Developer API pricing: Hydrogen's free and paid tier prices, grounding prices, the "Used to improve our products" row, and Carbon and Argon prices
- Google AI Studio, Gemini 4 Hydrogen model page: the model ID, input and output types, token limits and caching
- Google AI Studio, What's new in Gemini 4 Hydrogen: thinking levels, use cases, the Argon and Flash-Lite recommendations, the migration checklist, implicit caching and Managed Agents
- Google AI Studio, Choosing a model: Carbon's thinking tokens and use cases, the models on the free tier, and Carbon's and Argon's introductory price
- Google AI Studio, Gemini API release notes: the entry for Hydrogen's general availability
- Google DeepMind, Gemini 4 Hydrogen Model Card: the adaptation from Carbon, distribution channels, the Terminal-Bench 4.1 result, limitations and knowledge cutoff
- Google AI Studio, Gemini API Additional Terms of Service: data use on unpaid and paid services, human reviewers, the European Economic Area, Switzerland and United Kingdom exception, the definition of paid services, logging and storage
- Google AI Studio, Gemini API available regions: Türkiye
- Gemini Apps Help, About Gemini models: chat and task models by plan, auto model selection
- Google Cloud, Gemini Enterprise Agent Platform model list: generally available Gemini models
- Artificial Analysis, Gemini 4 Hydrogen analysis: Hydrogen's and Argon's Intelligence Index scores
