Google has launched Gemini 4 Hydrogen, the third model in its Gemini 4 family, and the price is the story: $1.50 per million input tokens and $7.50 per million output tokens, flat across the model’s full 1 million-token context window — with a free tier in the Gemini API, the first free tier for any Gemini 4 model. It went live worldwide on November 12, and unlike its two siblings, it is reaching every Google Search user, not just subscribers.
Hydrogen is not Google’s most powerful model. That title still belongs to Gemini 4 Carbon, which became generally available two days earlier. Hydrogen is the fast, cheap one: built for the routine work of software development and AI agents — writing code, running tests, fixing bugs, turning documents into reports — and designed to run all day without the token bill of a frontier model.
The price, and the catch
The headline rate is simple, and unusually so: one flat price, no surcharge for longer prompts. But the number needs context. At $1.50 and $7.50, Hydrogen costs twice what the model it replaces, Gemini 3.8 Flash, costs today. That sounds like a price rise until you look at the calendar: from January 1, 2027, Flash itself will cost $1.50 and $7.50. Hydrogen is not a cheaper Flash — it is Flash’s replacement at Flash’s future price, with far better performance attached.
Against the rest of the Gemini 4 family, Hydrogen is the cheapest option. Argon and Carbon both launched at $2 per million input tokens and $10 per million output tokens, and Carbon’s introductory rate doubles to $4 and $20 after November 30. Against Anthropic’s Claude Fable 5.2, launched on November 9 at $8 and $40, Hydrogen costs roughly a fifth. The trade-off is capability: Fable 5.2 still leads the broad Artificial Analysis index by a wide margin, and Carbon leads Hydrogen on it too, 60 points to 51.
One more number buyers should note: a free API tier sounds like free use, but free tiers come with rate limits that rarely suit production workloads. Developers who want to run Hydrogen at volume will pay — just less than they would for almost any rival of comparable quality.
Who can use it now
Hydrogen is the most widely available Gemini 4 model yet. Developers get it through the Gemini API and Google AI Studio on all tiers, including the free tier, and through the paid tier of Android Studio. Businesses can use it through Gemini for business, and paid Google AI Pro and Ultra subscribers get it in Antigravity, Google’s agent-first development platform. Consumers get it in the Gemini app and, for the first time with a Gemini 4 model, in Google Search for all users.
What it can do
Google’s claims for Hydrogen are striking, and they are all Google’s own numbers — no independent lab had reproduced them at launch. Against 3.8 Flash, the company says Hydrogen finishes tasks 2.5 times faster at its default setting, scores 57 percent against 18.2 percent on Terminal-Bench 4.1, and cuts fabricated or unsupported claims by more than five times.
The model also comes with four configurable thinking levels — low, medium, high and a new xhigh tier — so users choose how much reasoning a task gets and pay only for the depth it needs. At its default high level, Hydrogen scores 51 on the Artificial Analysis Intelligence Index, one point ahead of Argon; at xhigh it reaches 54. On specialist tests it leads its class on HLE-Diamond at 58.8 percent and ranks first on the Vals Tax Agent Bench, results that point to the professional, multi-step work Google is aiming it at.
For ordinary users, the most visible change will be in Search. Google’s Search team calls Hydrogen the biggest model upgrade to Search yet: better handling of multi-part and mistyped questions, more visual answers, and a wider range of relevant links. Free users of the Gemini app get Hydrogen through automatic model selection rather than a direct choice, so many people will meet the new model without ever seeing its name.
A crowded month at the frontier
Hydrogen is the third major model launch in ten days, after Anthropic’s Fable 5.2 on November 9 and Carbon’s general availability on November 10. The pattern is clear: the frontier labs are no longer competing only on who has the smartest model, but on who can put a capable model in front of the most people at the lowest price. For developers, the free API tier makes it costless to trial Hydrogen against the model they run today; for everyone else, no action is needed — the new model is already answering their searches.
