Google released Gemini 4 Hydrogen on November 12, 2026, and it landed in several places at once: the Gemini app, Google Search, the Gemini API, Google Antigravity and Gemini for business. That spread is why it matters beyond developers. If you publish anything online, the Search part is the one to watch.
The performance figures below come from Google unless a source is named. Independent testing has barely started, so I’ve flagged where the claims need checking.
What is Gemini 4 Hydrogen?
Gemini 4 Hydrogen is Google’s fast, cost-efficient Gemini 4 model for everyday coding, research and agent tasks. Google released it on November 12, 2026 as the upgrade from Gemini 3.8 Flash. It has a 1 million-token context window and sits below Gemini 4 Carbon and Gemini 4 Argon, which are built for long, deep reasoning.
Koray Kavukcuoglu, Google DeepMind’s SVP and Chief AI Architect, calls it the “daily driver for agentic workflows.” In plain terms, Carbon and Argon do the hard thinking, and Hydrogen does the fast, repeatable work: writing code, running tests, pulling information together.
Gemini 4 Hydrogen at a glance
| Detail | What Google says |
| Released | November 12, 2026 |
| Role | Fast execution model for everyday and agentic work |
| Replaces | Gemini 3.8 Flash (Google’s recommended upgrade) |
| Context window | 1 million tokens, flat pricing across the whole window |
| Thinking levels | Low, medium, high (default), xhigh |
| API price | $1.50 per million input tokens, $7.50 per million output tokens |
| Where to use it | Gemini app, Google Search, AI Studio, Gemini API, Android Studio, Gemini for business, Google Antigravity (paid AI Pro and Ultra) |
How fast and how smart is Gemini 4 Hydrogen?
Google says Hydrogen scores 51 on the Artificial Analysis Intelligence Index at its default high setting, just ahead of Argon’s 50. At the new xhigh setting it reaches 54. Against Gemini 3.8 Flash, Google reports tasks finishing 2.5 times faster at the default setting, and a Terminal-Bench 4.1 score of 57% versus 18.2%.
The Terminal-Bench jump is the number that stands out. Terminal-Bench tests how well a model finishes tasks using command-line tools, so tripling the score is a big claim for a model priced as a workhorse. Google also says Hydrogen cuts fabricated claims by more than 5x and recovers better when a tool call fails or a search comes back empty. It ranks fifth on arena.ai’s Agent Arena for multi-step tasks like deep research and data analysis.
On safety, Google reports a 96.7% score on its internal agent honesty evaluation and says Hydrogen is more than 2.3 times as resilient to indirect prompt injection (hidden instructions in web pages or documents) on the Gray Swan benchmark.
All of that is Google testing Google’s model. Treat it as a starting point.
How much does Gemini 4 Hydrogen cost?
Gemini 4 Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens through the Gemini API. The price stays flat across the full 1 million-token context window. Google AI Studio supports it on every tier, so you can test it before paying for anything.
The per-token price is only half the story. According to ImportStatic’s read of Google’s pricing page, Gemini 3.8 Flash costs $0.75 input and $3.75 output until December 31, so Hydrogen is double that for the rest of 2026. On January 1, 2027, 3.8 Flash moves to the same $1.50 and $7.50. Hydrogen also thinks at the high level by default, where 3.8 Flash defaulted to medium, so the same prompt can burn more output tokens than you’d expect.
Two more details from the same source matter if you’re building on the API. Hydrogen has no explicit context caching, only implicit caching, which bills cached input at 5% of the normal input price for prompts of at least 4,096 tokens. And the ImportStatic team suggests setting the thinking level yourself and starting at medium, so your first comparison with 3.8 Flash isn’t skewed by the new default.
The bigger pricing question is where this sits in the AI market. Cheap, fast models are now where the competition is, which fits the pattern in our piece on the AI price war turning into a fight over profitability.
How does Gemini 4 Hydrogen compare with Carbon and Argon?
Hydrogen is the fast, cheaper execution model. Carbon is Google’s strongest and uses the most thinking tokens. Argon handles advanced long-horizon work with fewer tokens than Carbon. Hydrogen beats Argon on one benchmark, but Google doesn’t pitch it as the smarter model overall.
| Model | Best for | Default thinking | Price per million tokens (input / output) | Released |
| Gemini 4 Hydrogen | Fast everyday coding and agent work | High | $1.50 / $7.50 | Nov 12, 2026 |
| Gemini 4 Argon | Long-horizon tasks with fewer tokens than Carbon | Medium | $2 / $10 now, $4 / $20 from Dec 1 | Oct 13, 2026 |
| Gemini 4 Carbon | The hardest reasoning problems | Medium | $2 / $10 now, $4 / $20 from Dec 1 | Nov 10, 2026 |
| Gemini 3.8 Flash | Previous everyday model | Medium | $0.75 / $3.75 until Dec 31 | Earlier |
Prices and defaults are from Google’s pricing and migration docs as summarized by ImportStatic on November 13, 2026. Check Google’s pricing page before you commit to a budget.
The names don’t tell you the ranking, which trips people up. Hydrogen is not a step above Carbon. It’s a different job.
What is Orchestrator in Google Antigravity?
Orchestrator is a model option in Google Antigravity that switches between Gemini 4 Carbon and Gemini 4 Hydrogen during one task. Carbon plans and diagnoses problems, Hydrogen writes edits and runs tests, and control passes back to Carbon when Hydrogen hits trouble. Google says it keeps near-Carbon success while using 24% less quota and finishing 19% faster.
The idea is a reaction to static model routers, which guess how hard a task is from your first prompt and send the whole thing to one model. Long agent runs don’t work that way. A task can be easy for twenty minutes and hard for two, and you often can’t tell which in advance. Orchestrator hands work back and forth as the task changes.
If you read those numbers closely, you’ll spot a mismatch. Google’s main announcement cites Terminal-Bench 4.0 for the Orchestrator result, while the Antigravity post says 4.1. It’s probably a typo in one place, but it’s worth knowing before you quote the figure.
Availability is split by plan. Hydrogen itself is open to paid Google AI Pro and Ultra subscribers in Antigravity. Orchestrator (Carbon) is for Ultra only. Orchestrator (Argon) is available to both Pro and Ultra. Developers who want to choose which model handles which role can wire Hydrogen and Carbon together through the Antigravity SDK, and Google shows an open-source issue triage demo built that way.
Who can use Gemini 4 Hydrogen today?
Google says the rollout is global, so it applies in India too. Business Standard’s New Delhi desk reported the same on November 13.
- Everyone: the Gemini app model menu and Google Search. Free Gemini app users get it through auto model selection when it fits the prompt.
- Developers: the Gemini API, Google AI Studio and Android Studio.
- Businesses: Gemini for business, plus Gemini in Google Sheets on eligible plans.
- Antigravity users: paid Google AI Pro and Ultra subscribers.
What does Gemini 4 Hydrogen change in Google Search and AI Mode?
Gemini 4 Hydrogen is now the default model in Google’s AI Mode for all users worldwide, and Google says it also handles the hardest AI Overview questions. Google says it improves query fan-out, the technique that runs many related searches at once, so answers should link to a wider range of websites.
This is the part that matters most for anyone who publishes content. Robby Stein, Google’s VP of Product for Search, said the fan-out upgrade shows more links to the web. Google’s Search post also promises more visual answers, with comparison tables, maps and custom tools, and better handling of multi-part questions and typos. Gemini 3.5 Flash had been AI Mode’s default since Google I/O in May, according to Search Engine Journal.
The timing is awkward for anyone tracking traffic. The switch came two days into Google’s November core update, which began November 10 and was still rolling out on November 12. Google hasn’t said which AI Overview questions count as hardest, and it hasn’t published any numbers for how many more links appear. For background on how AI Mode works, see our explainer on Google AI Mode and how the search giant is reinventing the internet’s front door.
What should site owners do about Gemini 4 Hydrogen?
Don’t rewrite your site this week. Do these four things instead.
- Log two dates: November 10 (core update begins) and November 12 (model change). Compare your traffic and AI citation data before and after each one, rather than reading November as a single event.
- Check your own numbers: Google says more links are coming, but gave no counts. Your citation and referral reports will show whether it applies to you. If you don’t track AI citations yet, our roundup of AI SEO tools is a place to start.
- Make each section answer one question: Fan-out splits a multi-part query into smaller searches. A page where every section opens with a direct, self-contained answer gives that system more to pick from. Google hasn’t published guidance linking the two, so treat this as reasoned inference, not a rule.
- Name sources, dates and people on the page: Claims that say who said what and when are easier for a person or a model to verify. It’s also the core of Google’s E-E-A-T approach, which we covered in why crediting original authors matters.
Where the Gemini 4 Hydrogen claims need checking
Google’s headline numbers are strong, but several come with caveats.
The benchmarks are Google’s own, run on tasks and settings Google chose. Independent reviews will take weeks. The Orchestrator result comes from a single Google test, and the version mismatch between Google’s two posts (4.0 versus 4.1) hasn’t been explained. The price is double Gemini 3.8 Flash’s until the end of the year, and the high default thinking level can raise real costs above the per-token math. And for Search, the open question is the one Google hasn’t answered: which AI Overview questions go to Hydrogen, and how many more links people actually see.
I’d watch the API bills and the citation data over the next few weeks before concluding. Our earlier look at Gemini 3.0 Pro and Google’s claim to have passed ChatGPT is a reminder that launch-week benchmarks and real-world results don’t always match.
Gemini 4 Hydrogen FAQ
What is Gemini 4 Hydrogen?
Gemini 4 Hydrogen is Google’s fast, cost-efficient Gemini 4 model for everyday coding, research and agent workflows. Google released it on November 12, 2026. It has a 1 million-token context window and is built for rapid execution, while Gemini 4 Carbon and Argon handle deeper, longer reasoning.
When was Gemini 4 Hydrogen released?
Google released Gemini 4 Hydrogen on November 12, 2026, across the Gemini app, Google Search, the Gemini API, Google AI Studio, Gemini for business and Google Antigravity. It followed Gemini 4 Carbon by two days and Gemini 4 Argon by about a month.
Is Gemini 4 Hydrogen free?
Partly. Free Gemini app users get Hydrogen through auto model selection when it suits their prompt, and it’s available in Google Search for everyone. Google AI Studio supports it on all tiers. Google Antigravity access requires a paid Google AI Pro or Ultra plan, and API use beyond free limits is billed per token.
Is Gemini 4 Hydrogen better than Gemini 4 Carbon?
No. Carbon is Google’s strongest model for deep, long-horizon reasoning. Hydrogen is built for speed and cost. Google reports Hydrogen scoring 51 on the Artificial Analysis Intelligence Index against Argon’s 50, but it positions Carbon and Argon for the hardest problems and Hydrogen for everyday execution.
How much does Gemini 4 Hydrogen cost per million tokens?
Gemini 4 Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens through the Gemini API, flat across its 1 million-token context window. Gemini 4 Argon and Carbon cost $2 and $10 now, rising to $4 and $20 from December 1, according to ImportStatic’s read of Google’s pricing page.
Does Gemini 4 Hydrogen power Google AI Mode?
Yes. Google made Gemini 4 Hydrogen the default model in AI Mode for all users worldwide on November 12, 2026, and says it also handles the hardest AI Overview questions. Paid Ultra subscribers can choose Gemini 4 Carbon in AI Mode for more complex work.
What does Gemini 4 Hydrogen replace?
Google recommends Hydrogen as the upgrade from Gemini 3.8 Flash for everyday coding and agent work. In AI Mode, Gemini 3.5 Flash had been the default since May. Gemini 3.8 Flash API pricing rises to $1.50 and $7.50 per million tokens on January 1, 2027, per ImportStatic.
