Quick Answer
Gemini 4 Hydrogen is Google’s fast-execution model in the Gemini 4 family, launched November 12, 2026. It is adapted from Gemini 4 Carbon but tuned for speed and efficiency on everyday coding and agentic workflows rather than the deepest reasoning. It accepts text, images, audio and video, has a 1-million-token context window with up to 1 million output tokens, and offers four thinking levels — low, medium, high (default) and a new xhigh tier. It costs $1.50 per million input tokens and $7.50 per million output tokens, flat across the whole context window, with a rate-limited free tier in Google AI Studio and the Gemini API. It is rolling out in the Gemini API, AI Studio, the Gemini app, Google Search, Gemini for business and Google Antigravity.
What Is Gemini 4 Hydrogen?
Every model family needs a fast tier. In the Gemini 4 family, Carbon and Argon handle deep architectural judgment and extended reasoning; Hydrogen handles the work that has to happen quickly and repeatedly. Google describes it as the execution model: it writes code, runs tests, verifies changes, does research and drives tool-using agents, while the bigger models are reserved for the harder planning stages.
It is adapted from Gemini 4 Carbon, so it inherits the family’s multimodal understanding, but its design point is different — finish the task fast, use less quota doing it, and keep the quality high enough that the result is usable without a frontier model checking it. Google positions it for “everyday coding and agentic workflows,” and in the Gemini app and Google Search it is the model free users now meet through automatic model selection.
Hydrogen also replaces Gemini 3.8 Flash across Google’s products. That matters for anyone who built on Flash: the fast tier is no longer a 3.8-generation model, and the numbers Google reports against Flash — 2.5x faster, Terminal-Bench 4.1 up from 18.2% to 57% — are the size of the upgrade.
Release Date and Availability
Google announced Gemini 4 Hydrogen on November 12, 2026, in a post on blog.google by Koray Kavukcuoglu, SVP at Google DeepMind and Chief AI Architect at Google, titled “Introducing Gemini 4 Hydrogen for Agentic Workflows.” It arrived two days after Gemini 4 Carbon went generally available and one day after the Gemini Decisions API launched — a busy week for the Gemini family.
It is available now, rolling out globally across:
- Gemini API and Google AI Studio — all tiers, including a rate-limited free tier
- Gemini app — free users receive it through automatic model selection when Google considers it appropriate for the prompt
- Google Search — better handling of multipart or typo-filled queries, more visual responses and expanded query fan-out
- Gemini for business — enterprise access
- Google Antigravity — for paid Google AI Pro and Ultra subscribers
- Gemini in Google Sheets — for Google AI Pro and Ultra subscribers
Its knowledge cutoff is March 2026, which may make it less current in fast-moving domains.
Key Features
1. Built for speed on agentic work
Hydrogen’s job is execution: repository coding and debugging, interactive web development, research, data analysis and broad codebase navigation. Google reports it finishes tasks 2.5 times faster than Gemini 3.8 Flash at its default high thinking setting.
2. Four thinking levels, including a new xhigh tier
Hydrogen offers low, medium, high (the default) and a new xhigh level. Google recommends high for most complex tasks and reserving xhigh for harder multi-step problems where high falls short, noting that xhigh can overthink simpler prompts. The levels scale quality, latency and cost in a smooth progression, so you pay for only the reasoning depth a task needs.
3. 1 million tokens in, up to 1 million tokens out
Hydrogen pairs a native 1-million-token input window with up to 1 million output tokens, sharing one 1,048,576-token budget across input, thinking and output. That lets it synthesize dense PDFs, spreadsheets, charts, audio and recorded video into polished, cited reports and structured deliverables. Note the trade-off: reasoning consumes capacity that could otherwise hold input or final output, and higher thinking settings also increase latency.
4. Native multimodal input
It accepts text, images, audio and video. In the Gemini app it follows tone and formatting instructions more closely, draws on uploads and connected services including Gmail, Drive, Airtable and Dropbox, and selects tools more reliably, according to Google.
5. Antigravity Orchestrator — Carbon plans, Hydrogen executes
Google also shipped Orchestrator inside Google Antigravity, a new option that splits planning from execution within a single session. Carbon explores dependencies and produces a plan in a read-only phase; Hydrogen then edits code, updates packages and runs tests. If execution hits friction, control returns to Carbon to inspect the state and repair the plan. Unlike a static router that commits a whole task to one model, Orchestrator can change models as the work changes. Google reports near-Carbon task success on Terminal-Bench 4.0 while using 24% less quota and finishing 19% faster on average. Orchestrator initially rolls out to AI Ultra subscribers, with enterprise access through Gemini for business.
6. Safer agent behaviour than Flash
Google reports Hydrogen cuts agentic overreach by up to 81% and under-service by up to 79% versus Gemini 3.8 Flash, and is more than 2.3 times more resilient to indirect prompt injection on the Gray Swan IPI benchmark. Those measures address the failure modes that matter when a model can select tools and act across connected services.
Pricing Explained
Google lists one flat paid rate: $1.50 per million input tokens and $7.50 per million output tokens, flat across the entire 1-million-token context window — long prompts do not trigger a higher tier.
| Tier | Price |
|---|---|
| Paid (Gemini API) | $1.50 per 1M input tokens / $7.50 per 1M output tokens |
| Cached input (implicit caching) | $0.075 per 1M tokens |
| Free tier (Google AI Studio / Gemini API) | $0 — rate-limited |
| Gemini app (consumers) | Included — Hydrogen is served through automatic model selection |
One secondary analysis notes Hydrogen costs twice what Gemini 3.8 Flash costs today, and the same as Flash will cost from January 1, 2027 — so the price rise is real, but it arrives with a much faster model, and the comparison is against Flash’s outgoing rate.
There is no separate price for the thinking levels: the same $1.50/$7.50 rate applies at low, medium, high and xhigh. What changes with the level is how many tokens the model spends thinking — and thinking tokens share the same 1-million-token window as input and output.
Gemini 4 Hydrogen vs Carbon, Argon and 3.8 Flash
| Feature | Gemini 4 Hydrogen | Gemini 4 Carbon | Gemini 4 Argon | Gemini 3.8 Flash |
|---|---|---|---|---|
| Role | Fast execution, everyday agentic work | Deep reasoning, frontier tasks | Deepest reasoning, hardest problems | Previous fast tier (being replaced) |
| Context window | 1M tokens in, up to 1M out | 1M tokens | 1M tokens | Smaller (previous generation) |
| Thinking levels | Low, medium, high (default), xhigh | 3 levels (low/medium/high) | 3 levels (low/medium/high) | — |
| Price (input / output per 1M) | $1.50 / $7.50 | $2 / $10 until 30 Nov, then $4 / $20 | Higher (frontier tier) | Lower (outgoing rate) |
| Intelligence Index | 51 at high, 54 at xhigh | Higher on deep-reasoning benchmarks | Higher on deep-reasoning benchmarks | Lower |
| Terminal-Bench 4.1 | 57% | Higher | Higher | 18.2% |
| Best for | Coding, agents, everyday tasks | Hard reasoning, long tasks | Hardest problems | Replaced by Hydrogen |
Hydrogen is not the smartest Gemini 4 model — Carbon and Argon keep that job. Its edge is speed per dollar on the work most developers actually run every day. Google’s own headline comparison makes the point: Hydrogen scores 51 on the Artificial Analysis Intelligence Index at its default high setting against Argon’s 50, rising to 54 at xhigh — a thin lead on that one index, measured at Hydrogen’s default against Argon’s default configuration, and not a claim that Hydrogen out-reasons Argon overall. If you want the wider Gemini family first, see our What Is Gemini 4 Carbon? overview and our What Is Google Gemini and How Does It Work? explainer.
How to Use Gemini 4 Hydrogen
1. Try it free in the Gemini app
Free users receive Hydrogen through automatic model selection — Google decides when a prompt fits it. There is no manual switch to claim; ask a coding or research question and the app may serve Hydrogen behind the scenes.
2. Use it in Google AI Studio
Sign in at aistudio.google.com, where Hydrogen is available on all tiers including the rate-limited free tier. Select Gemini 4 Hydrogen as the model and set the thinking level — high for most complex tasks, xhigh only for harder multi-step problems where high falls short.
3. Call it from the Gemini API
Hydrogen is available through the Gemini API for developers building coding assistants and tool-using agents. It accepts text, images, audio and video input and returns text. Synchronous responses support up to 65,536 tokens per hop, while background execution or client-side continuation can extend generation up to the remaining capacity in the shared 1-million-token window.
4. Use it in Antigravity with Orchestrator
In Google Antigravity (paid Pro and Ultra subscribers), Hydrogen is the execution engine: Carbon plans in a read-only phase, Hydrogen edits code, updates packages and runs tests, and control returns to Carbon when the plan needs repair. Orchestrator initially rolls out to AI Ultra subscribers.
5. Watch the thinking budget
Because input, thinking and output share one 1,048,576-token window, a high or xhigh setting on a long input leaves less room for the answer. For long documents, start at high and raise the level only when the result falls short.
Best Uses for Gemini 4 Hydrogen
- Everyday coding: repository coding and debugging, interactive web development, writing code and running tests in a loop.
- Agentic workflows: multi-step tasks like deep research, data analysis and broad codebase navigation, where the model selects tools and acts.
- Long-document synthesis: turning dense PDFs, spreadsheets, charts, audio and recorded video into cited reports and structured deliverables.
- Search and everyday questions: the Gemini app and Google Search now serve Hydrogen for suitable prompts, including multipart or typo-filled queries.
- Multi-model pipelines: use Carbon to plan and Hydrogen to execute — the Orchestrator pattern in Antigravity — so frontier reasoning is spent only on the hard stages.
- Prototyping on the free tier: the rate-limited free tier in AI Studio and the API makes it a cheap starting point for agent experiments.
Limitations and Things to Know
- It is not the frontier model. Carbon and Argon out-reason Hydrogen on deep problems. The 51-vs-50 Intelligence Index edge over Argon is one index at default settings, not an overall superiority claim.
- A rival still leads on the cheating-adjusted coding table. The model card’s cheating-adjusted Terminal-Bench 4.1 table lists Claude Sonnet 5.5 at 63.9%, above Hydrogen’s 57.0%. Hydrogen improves sharply on Google’s previous Flash model without leading every competitor in the same coding comparison.
- Unjustified refusals rose. The model card records a 5.57-percentage-point increase in unjustified refusals versus Gemini 3.8 Flash — the model says no more often when it should not.
- Benchmarks are Google’s own. Every figure in this article comes from Google’s announcement or its model card, on benchmarks Google chose and measured. No independent evaluation exists yet.
- Known limitations in the model card. Hallucinations, occasional slowness or timeouts, and potentially higher token use at elevated thinking levels.
- Knowledge cutoff is March 2026. It may be less current in fast-moving domains.
- Safety assessment is mixed. The frontier-safety assessment found no material general-capability increase over Argon, and targeted tests placed Hydrogen below the biological-information and cybersecurity critical capability levels — but it reaches the alert thresholds in both domains, and fuller safety documentation is still forthcoming.
- xhigh can overthink. Google warns that the highest thinking level can overthink simpler prompts, costing time and tokens for no gain.
Frequently Asked Questions
What is Gemini 4 Hydrogen?
Google’s fast-execution model in the Gemini 4 family, launched November 12, 2026. It is built for everyday coding and agentic workflows — writing code, running tests, doing research and carrying multi-step tasks to completion — and it replaces Gemini 3.8 Flash as Google’s fast tier.
How much does Gemini 4 Hydrogen cost?
$1.50 per million input tokens and $7.50 per million output tokens, flat across its full 1-million-token context window. Cached input costs $0.075 per million tokens through implicit caching, and there is a rate-limited free tier in Google AI Studio and the Gemini API.
Is Gemini 4 Hydrogen free?
There is a rate-limited free tier through the Gemini API and Google AI Studio, and free Gemini app users receive Hydrogen through automatic model selection. Paid API use starts at $1.50/$7.50 per million tokens.
How is Gemini 4 Hydrogen different from Gemini 4 Carbon?
Carbon is the deep-reasoning frontier model; Hydrogen is adapted from Carbon but tuned for speed and efficiency on everyday work. Hydrogen is faster and cheaper per token, while Carbon handles the harder planning and reasoning stages — in Antigravity’s Orchestrator, Carbon plans and Hydrogen executes.
What are the thinking levels in Gemini 4 Hydrogen?
Four: low, medium, high (the default) and a new xhigh tier. Google recommends high for most complex tasks and xhigh only for harder multi-step problems, since xhigh can overthink simple prompts.
How fast is Gemini 4 Hydrogen?
Google says it finishes tasks 2.5 times faster than Gemini 3.8 Flash at its default high setting, and scores 57% on Terminal-Bench 4.1 against Flash’s 18.2%. Those are Google’s own figures; independent verification has not been published yet.
Where can I use Gemini 4 Hydrogen?
In the Gemini API, Google AI Studio (all tiers), the Gemini app, Google Search, Gemini for business, Google Antigravity (Pro and Ultra subscribers) and Gemini in Google Sheets (Pro and Ultra subscribers).
Does Gemini 4 Hydrogen replace Gemini 3.8 Flash?
Yes. Google says Hydrogen has taken the place of 3.8 Flash across its products, and it is the first Gemini 4 model served to free users in the Gemini app.
Final Verdict
Gemini 4 Hydrogen is the least glamorous and most useful kind of Gemini 4 launch: not a new frontier, but a faster daily driver. For the work developers actually run every day — coding, testing, research, multi-step agent tasks — 2.5x the speed of Flash and a Terminal-Bench jump from 18.2% to 57% is a real upgrade, and the flat $1.50/$7.50 pricing makes the bill predictable. The honest limits are clear too: it does not out-reason Carbon or Argon, a rival still leads the cheating-adjusted coding table, and it refuses more often than Flash did. Use it as the execution engine — with Carbon planning the hard stages where that matters — and judge it on your own tasks, not on Google’s benchmarks. Start on the free tier in AI Studio, keep the thinking level at high unless a task genuinely needs xhigh, and watch the shared token window on long inputs.
Editorial update: November 13, 2026. This article is based on Google’s November 12, 2026 announcement and the Gemini 4 Hydrogen model card, checked November 13, 2026.