Google launches Gemini 4 Hydrogen, its fast execution model, across Search, the app and the API
Google DeepMind launched Gemini 4 Hydrogen on November 12, the fast, execution-oriented member of the Gemini 4 family, rolling out globally across Google AI Studio, the Gemini API, Gemini for business, the Gemini app and Google Search for all users, and in Antigravity for paid AI Pro and Ultra subscribers. Google prices it flat at $1.50 per million input tokens and $7.50 per million output tokens across a 1-million-token context window, and says it scores 51 on the Artificial Analysis Intelligence Index at its default setting, against 50 for Argon.
📊 Scorecard: our Bullish call on $GOOGL is scored from $298.70, the last close before it ran. Its first result comes after the next close.
The News
Google DeepMind launched Gemini 4 Hydrogen on November 12, the fast, execution-oriented member of the Gemini 4 family, rolling out globally across Google AI Studio, the Gemini API, Gemini for business, the Gemini app and Google Search for all users, and in Antigravity for paid AI Pro and Ultra subscribers. It is adapted from the frontier Gemini 4 Carbon and built for rapid execution across everyday coding and agentic workflows — work where an AI agent carries out multi-step tasks on a user's behalf — rather than deep, long-horizon reasoning, which stays with Carbon and Argon.
Google prices Hydrogen flat across its 1-million-token context window at $1.50 per million input tokens and $7.50 per million output tokens, and says it has four thinking levels: low, medium, high as the default, and a new xhigh. A token is a chunk of text, roughly three-quarters of a word, so a 1-million-token context window means the model can hold a very large document or a long conversation in mind at once. Flat pricing means the price does not rise as the conversation gets longer — the same rate applies across the whole window.
Google's own figures put Hydrogen at 51 on the Artificial Analysis Intelligence Index at the default high setting, versus 50 for Argon, reaching 54 at xhigh. Google also says tasks finished 2.5 times faster than Gemini 3.8 Flash, and Terminal-Bench 4.1 — a test of how well a model completes real coding tasks in a terminal — up to 57%, from 18.2% for 3.8 Flash. Those are Google's own figures and have not been independently reproduced. One independent check, by orcarouter.ai on November 13, found the 2.5-times speed claim does not survive the same outside harness, while confirming the index score — so the speed claim should be read as Google's, not as established fact. The knowledge cutoff is March 2026.
Hydrogen is Google's bet that the next AI race is won on speed of execution, not depth of thought.
Why It Matters
The model family now has a division of labour, and Hydrogen's job is speed. Carbon and Argon keep the deep, long-horizon reasoning — the hard problems where a model needs to think for a long time before answering. Hydrogen is for the opposite kind of work: everyday coding, and agentic workflows where an AI agent carries out multi-step tasks, calling tools, writing code and checking its own work. Those jobs are numerous and repetitive, and for them the model that finishes first, at a predictable price, wins.
The pricing is the quiet part of the launch. Flat pricing across a 1-million-token context window means a developer pays the same $1.50 per million input tokens and $7.50 per million output tokens whether the conversation is short or very long. Most models get more expensive, per token, as the context grows — because holding more in mind costs more compute. Google is absorbing that cost, which makes Hydrogen easy to budget for and easy to leave running on long agent tasks. That is a pricing decision aimed at volume.
The benchmark numbers need the same care as the price. An intelligence index score of 51 against Argon's 50, and 54 at the new xhigh thinking level, are Google's own figures, and Terminal-Bench 4.1 at up to 57% against 18.2% for 3.8 Flash is a large jump on a test that measures real coding work in a terminal. But benchmarks are chosen by the vendor, and the first independent check — orcarouter.ai on November 13 — found the 2.5-times speed claim does not survive the same outside harness, while confirming the index score. The honest reading is that Hydrogen is genuinely fast and capable, and that the exact speed advantage is still contested.
For Google, the launch is also about distribution. Rolling Hydrogen out across Search, the Gemini app, the API, AI Studio, Gemini for business and Antigravity at once means it reaches every kind of user on day one — from a person asking Search a question to a developer building an agent. That breadth is what turns a model launch into a product advantage: the model does not have to be found, it is already where the users are.
Analyst's Take
Our call is Bullish. Hydrogen is a deliberately practical launch: a fast model priced flat at $1.50 per million input tokens and $7.50 per million output tokens, rolled out across Google's main surfaces at once (Antigravity for paid Pro/Ultra subscribers), aimed squarely at the everyday coding and agentic work where developers actually spend their money. Google's own figures put it ahead of Argon on the intelligence index at its default setting, and the Terminal-Bench jump from 18.2% to 57% against 3.8 Flash, if it holds, is a large step for a model built for speed rather than depth.
What to watch: whether independent testing confirms the speed and benchmark claims — the orcarouter.ai check already disputes the 2.5-times speed figure while confirming the index score, so treat the headline numbers as Google's until more outside harnesses weigh in. And watch adoption: Hydrogen's value to Google is not a leaderboard score, it is developers choosing it as their daily driver inside Search, the Gemini app and the API. If usage follows the rollout, the flat pricing becomes a volume story.
Are you bullish or bearish on this story?
By The Numbers
FAQ
What is Gemini 4 Hydrogen?
It is the fast, execution-oriented member of Google's Gemini 4 model family, launched November 12. It is adapted from the frontier Gemini 4 Carbon and built for rapid execution across everyday coding and agentic workflows — work where an AI agent carries out multi-step tasks — rather than deep, long-horizon reasoning, which stays with Carbon and Argon.
Where can you use it?
It is rolling out globally across Google AI Studio, the Gemini API, Gemini for business, the Gemini app and Google Search for all users, and in Antigravity for paid AI Pro and Ultra subscribers.
How much does it cost?
Google prices it flat across its 1-million-token context window: $1.50 per million input tokens and $7.50 per million output tokens. A token is a chunk of text — roughly three-quarters of a word — so a million tokens is a very large document or conversation.
What are the thinking levels?
Google says Hydrogen has four thinking levels — low, medium, high as the default, and a new xhigh. Thinking levels control how much reasoning the model does before answering: more thinking usually means better answers on hard tasks, but slower ones.
What do the benchmark numbers mean?
Google's own figures put Hydrogen at 51 on the Artificial Analysis Intelligence Index at the default high setting, versus 50 for Argon, reaching 54 at xhigh; Terminal-Bench 4.1 up to 57%, from 18.2% for Gemini 3.8 Flash. These are Google's own figures and have not been independently reproduced — and one independent check found the 2.5-times speed claim does not hold up outside Google's harness.
👁
Source: Google Keyword Blog