Google launched Gemini 4 Hydrogen on November 12, 2026. It is the fast, everyday-work model in the Gemini 4 family: built for rapid execution across coding and agentic workflows, rather than the deep, long-horizon reasoning that the frontier models handle. It is already available in AI Studio, the Gemini API, the Gemini app, Google Search, and Google Antigravity — and it is the model most people will actually use day to day.
This guide explains what Hydrogen is, how it differs from Gemini 4 Carbon and Argon, what it costs, and when you should pick it over the bigger models. If you write code with AI, run agents, or just use Gemini in Search, Hydrogen is the model doing more of that work now — and the differences from the previous generation are large enough to matter.
What Is Gemini 4 Hydrogen?
Gemini 4 Hydrogen is Google's execution model. While the frontier models in the same family — Gemini 4 Carbon and Gemini 4 Argon — are built for deep reasoning on hard problems, Hydrogen is built for speed and efficiency on everyday work. Google describes it as the model for rapid execution across diverse agentic workflows: writing code, running tests, verifying changes, and carrying out multi-step tasks without waiting on a heavyweight model.
The important shift is that Hydrogen is not a cut-down model. It is a full Gemini 4 model with a 1 million-token context window, available across all of Google's surfaces at once. Developers get it in AI Studio and the Gemini API. Businesses get it through Gemini for business. Everyone gets it in the Gemini app and in Google Search. Paid Google AI Pro and Ultra subscribers also get it in Google Antigravity, Google's agent-first development environment.
That wide rollout is the point. Google is putting the same fast model behind the coding assistant, the API, the consumer app, and Search — so the model you meet in one place behaves like the model you meet in the others.
For comparison, the last big model launch covered here was Claude Fable 5.2, which arrived as one model with two safety boundaries; Hydrogen arrives instead as one tier of a three-model family, with a defined job inside it.
Hydrogen vs Carbon vs Argon: The Three Gemini 4 Models
Think of the Gemini 4 family as a team, not a ladder. Carbon and Argon do the deep thinking: architecture decisions, untangling subtle bugs, long-horizon planning. Hydrogen does the doing: writing the code, running the tests, and verifying the result. The models are designed to work together, not to compete with each other — and Google built a tool, Orchestrator, specifically to switch between them inside one session.
| Model | Role | Best for |
|---|---|---|
| Gemini 4 Carbon | Frontier reasoning | Hard problems, architecture, long-horizon planning |
| Gemini 4 Argon | Frontier reasoning | Deep analysis and complex reasoning tasks |
| Gemini 4 Hydrogen | Fast execution | Everyday coding, agents, Search, routine work |
On benchmarks, Hydrogen holds its own against its bigger siblings. It surpasses Argon on the Artificial Analysis Intelligence Index at its default high thinking level — 51 against 50 — and reaches 54 at its highest xhigh setting, according to Artificial Analysis. That is one index, not an overall verdict, but it shows Hydrogen is not giving up much intelligence for its speed.
It also finishes tasks faster than models in its class, which is the trade that matters for everyday work.
What Hydrogen Is Actually Good At
Hydrogen is a major upgrade over Gemini 3.8 Flash, the fast model it replaces. Google says it finishes tasks 2.5 times faster at its high default setting, while more than tripling its Terminal-Bench 4.1 score — 57% against 18.2% — and cutting fabricated claims by more than five times. It also adapts cleanly when a tool call fails or a search returns nothing, which matters more in real agent work than any single benchmark number.
For developers, the strengths are specific. In iterative repository coding and debugging, Hydrogen makes focused, surgical edits instead of rewriting unrelated files. It batches file-read and search calls in parallel rather than inspecting files one by one.
In early testing across complex graphs, SQL, and data-heavy enterprise workflows, Google's partners saw it run two to 2.8 times faster while using 12% to 23% fewer tokens than 3.8 Flash. On multi-step agentic work — deep research, data analysis, broad codebase navigation — it ranks fifth on Agent Arena's live benchmark.
For knowledge work, Hydrogen pairs its 1 million-token input with up to 1 million output tokens. It can take dense PDFs, spreadsheets, charts, audio, and recorded video and turn them into cited reports and structured deliverables.
It leads its class on HLE-Diamond at 58.8%, ranks first overall on the Vals Tax Agent Bench at 52.3%, ties for first on the Vals Legal Research Bench at 55.8%, and leads its class on AutomationBench-AA. On the Multimodal Knowledge Agent benchmark it scores 53.4%, and on Agents' Last Exam v1.1 it scores 71.6%.
For web development, Hydrogen generates more functional, feature-complete layouts in fewer prompts. That is the use case most readers will notice first: live UI prototyping, design-to-code iteration, and interactive app building, where speed per prompt matters more than exhaustive reasoning. The model is tuned to get to a working result quickly and then refine it, rather than planning at length before producing anything.
Gemini 4 Hydrogen Pricing: What It Costs to Use
Gemini 4 Hydrogen is priced flat across its entire 1 million-token context window: $1.50 per million input tokens and $7.50 per million output tokens. There is no price change for longer contexts, which makes costs predictable: a task that reads a lot and writes a little stays cheap, and the expensive direction is always output, not input.
That flat pricing is part of the model's pitch for multi-model workflows. The expensive frontier models handle the planning, and Hydrogen handles the execution — the part of the job that generates the most tokens. Because Hydrogen is cheaper per token and faster per task, the combined workflow costs less than running a frontier model end to end, while keeping the frontier model's judgment where it matters.
For consumers, there is no separate price to pay: Hydrogen is the model behind the Gemini app and Google Search for all users. Developers pay through the Gemini API in AI Studio, and businesses through Gemini for business. Paid Google AI Pro and Ultra subscribers also get Hydrogen inside Google Antigravity. If you already use any of those surfaces, Hydrogen is there now — there is nothing to install or enable.
The Four Thinking Levels: Low, Medium, High, XHigh
Hydrogen comes with four thinking levels: low, medium, high, and a new xhigh tier. High is the default. The levels trade reasoning depth against latency and cost, so you only wait — and pay — for the thinking a task actually needs. A quick rewrite can run at low; a tricky debugging session can run at xhigh.
The progression is smooth by design. Across the levels, Hydrogen matches or outperforms competing models in its class while often finishing faster and using fewer output tokens. The xhigh setting is where the headline benchmark numbers come from: 54 on the Artificial Analysis Intelligence Index, against Argon's 50. At the default high setting it scores 51 — still ahead of Argon on that index, at a lower cost in time and tokens.
For most users this choice will be invisible: the Gemini app and Search pick a level for you. Developers using the API can set it explicitly, and that is where it matters. If your agent loops are slow or expensive, dropping the thinking level is the first lever to try — and if the output quality drops, raising it back is a one-line change.
Orchestrator: When One Model Is Not Enough
Google also introduced Orchestrator in Google Antigravity. Instead of using one model from start to finish, Orchestrator switches between models like Carbon and Hydrogen inside the same session. The frontier model plans; Hydrogen executes. For parallel work, Carbon can act as the coordinator while Hydrogen powers worker and verifier agents that build and test changes at the same time in isolated workspaces.
The numbers Google reports are concrete. On Terminal-Bench 4.0, Orchestrator reached near-Carbon task success rates while consuming 24% less quota and completing tasks 19% faster on average. That is the whole argument for the multi-model approach in one line: nearly the same result, for less money, in less time.
Orchestrator is available to paid Google AI Pro and Ultra subscribers in Antigravity. If you are not on those plans, the idea still applies to your own workflows: use the strongest model you have for planning and the fastest model you have for doing. You do not need Google's tool to follow that pattern — but Orchestrator is the first time Google has automated the switching for you.
What Hydrogen Changes in Google Search
Hydrogen is also upgrading Google Search for all users globally, and this is where most people will meet it. Google says Search now understands your true intent better — including multi-part questions, typos, and very specific requests — and gives more visual, easy-to-scan responses. It is also upgrading the query fan-out technique, which connects you to a broader range of helpful links than before.
That last point matters for websites. A wider fan-out means Search can surface sites it previously might not have shown for a query. For site owners, the November core update is still rolling out at the same time — see Google's November AI Mode updates for what else changed this month — so ranking movement right now has more than one cause. Do not credit Hydrogen, or blame it, for a single week's movement.
In the Gemini app, Hydrogen follows instructions more closely across multi-turn conversations and calls the right tools more often. It pulls context from uploads, Gmail, Drive, and connected apps like Airtable and Dropbox, and it sticks to your tone and format on writing tasks. The honest summary: the assistant gets less confused, asks fewer clarifying questions, and produces usable output on the first try more often.
Should You Switch to Hydrogen?
If you use Gemini, you have already switched — Hydrogen is the model behind the app and Search now. The real question is for developers and teams choosing models deliberately. Use Hydrogen as your default for coding, testing, and agent work, and bring in Carbon or Argon when a task needs deep judgment: architecture, a subtle bug, or a decision with real consequences.
Watch two things as you adopt it. First, the safety numbers: Hydrogen reduces overreach in agentic tasks by up to 81% and under-service by up to 79%, cuts misaligned actions and task shortcutting by 12 times, and is more than 2.3 times more resilient to indirect prompt injection attacks on the Gray Swan benchmark.
It also preserves 99.9% chain-of-thought legibility with zero deceptive reasoning edits across more than 200,000 audited turns. Those are Google's own evaluations, but the direction is clear: the fast model is also the more disciplined one.
Second, keep an eye on the privacy side. Hydrogen pulls context from Gmail and Drive when you ask it to, which is powerful and worth understanding. Gemini's human reviewers can read personal files in some circumstances, so know what you are sharing before you connect an app. The model is fast and capable; the judgment about what to feed it is still yours.
Hydrogen is not the most powerful model Google makes, and it does not try to be. It is the model for the work you do all day: fast, cheap enough to use freely, and good enough that the frontier models can stay on the shelf until a task truly needs them. That is a useful thing for Google to have built — and a useful thing to understand before you pick a model for your next project.
Frequently Asked Questions
What is Gemini 4 Hydrogen?
Gemini 4 Hydrogen is Google's fast execution model in the Gemini 4 family, launched November 12, 2026. It is built for rapid execution across coding and agentic workflows, with a 1 million-token context window. It sits alongside the frontier reasoning models Carbon and Argon, which handle deep, long-horizon reasoning.
Is Gemini 4 Hydrogen free?
For consumers, yes in practice: Hydrogen powers the Gemini app and Google Search for all users at no extra charge. Developers pay through the Gemini API at $1.50 per million input tokens and $7.50 per million output tokens. Paid Google AI Pro and Ultra subscribers also get it in Google Antigravity.
How much does Gemini 4 Hydrogen cost?
The API price is flat across the full 1 million-token context window: $1.50 per million input tokens and $7.50 per million output tokens. There is no surcharge for longer contexts. Output tokens cost five times input tokens, so tasks that read a lot and write a little stay cheap.
Is Gemini 4 Hydrogen better than Argon?
It depends on the task. Hydrogen surpasses Argon on the Artificial Analysis Intelligence Index — 51 against 50 at its default high thinking level, and 54 at xhigh. But Argon is a frontier reasoning model built for deep analysis, so for hard reasoning tasks Argon is still the stronger choice. Hydrogen wins on speed, cost, and everyday execution.
What is the difference between Hydrogen and Gemini 3.8 Flash?
Hydrogen replaces 3.8 Flash as Google's fast model. It finishes tasks 2.5 times faster at its high default setting, scores 57% on Terminal-Bench 4.1 against 18.2% for 3.8 Flash, and cuts fabricated claims by more than five times. It also handles failed tool calls and empty searches more gracefully.
Where can I use Gemini 4 Hydrogen?
Hydrogen is available in AI Studio, the Gemini API, Gemini for business, the Gemini app, Google Search, and Google Antigravity for paid Google AI Pro and Ultra subscribers. It rolled out globally on November 12, 2026, so if you use any of those surfaces, it is already there.
