Google has started naming its machines after the periodic table. First came Carbon, then Argon, and now, on 12 November 2026, Hydrogen, the first element in the table and the lightest. As Ryan Whitwam wrote for Ars Technica, the new names tell you less than the old Pro and Flash labels did. Hydrogen’s job is still easy to read. It takes over from the Flash line: the faster, cheaper models that most people in Google’s ecosystem actually use, often without knowing it.

So this is not the model that sits on the summit. It is the one that does the daily work.

The order of the queue

Each of the big releases came with a queue. Google tested Carbon with its own staff before anything was announced. When Carbon and Argon did arrive, they went to cybersecurity researchers first. Even after general availability, Google kept them behind its most expensive AI subscription. Argon went on sale in October, and the front of the line was reserved for people who could pay for it.

White figures queue behind dark ropes on a blue background while a gold figure carrying a briefcase walks toward a large blank microchip.

Hydrogen got no such ceremony. Access began immediately, and the model has replaced Gemini 3.8 Flash across Google’s products. According to Google’s announcement, it is rolling out worldwide to developers through the Gemini API, AI Studio and Android Studio, to businesses through Gemini for business, and to consumers through the Gemini app and Google Search. Free users of the Gemini app have no menu to choose it from. They get Hydrogen through auto model selection, when the system decides it is the right model for the prompt. In Google Antigravity, the company’s coding environment, it is available only to paid AI Pro and Ultra subscribers.

Search is the biggest prize. In a separate post on Search, Google said Hydrogen is now available there to all users worldwide. PPC Land pointed out that this came thirty days after Argon, the first Gemini 4 model in Search, was put there for paying subscribers only. The timing is awkward. The same week, Google’s folding of AI into Search drew a lawsuit and protests in seven cities.

A reflection, and only that: a powerful institution rations its prized possession at first, keeping it for the trusted and the wealthy. What it hands out freely is the cheaper version, and the cheaper version is the one that ends up shaping everyone’s day.

A large watering can pours streams of water into thirteen cups against a blue background, beside a gold trophy under a glass dome.

The company grades its own paper

Google says Hydrogen is a big step up from the model it replaces. It can run at four thinking levels: low, medium, high and a new one called xhigh. At high, which Google’s announcement describes as the default, the company says it is 2.5 times faster than 3.8 Flash. Google’s own analysis says it makes more than five times fewer “fabricated claims”. On coding, Hydrogen scored 57 percent on Terminal-Bench 4.1, according to the company, against 18.2 percent for 3.8 Flash. That is more than triple.

The awkward number is the one that compares Hydrogen with its bigger sibling. On the Artificial Analysis Intelligence Index, a widely watched ranking, Google reports that Hydrogen at xhigh scored 54. Argon scored 50. The company’s post adds that Hydrogen scores 51 even at its default setting, which still puts it above Argon. On this one index, then, the cheap model beats the expensive model that was rationed only weeks earlier. Google makes no such claim across the board: Hydrogen goes toe-to-toe with Argon on some tests, not all of them.

The post is signed by Koray Kavukcuoglu, Google DeepMind’s senior vice president and the company’s chief AI architect. He calls Hydrogen “our daily driver for agentic workflows”, meaning software that carries out tasks rather than just answering questions. The safety figures are about exactly that. Google says that compared with 3.8 Flash, Hydrogen cuts overreach in agentic tasks by up to 81 percent. It also says the model is up to two times less likely to get around permission limits when a task is blocked or impossible.

A cream robotic arm grasps a yellow star above a dark navy machine against a bright blue background.

Almost all of these numbers come from the company that built the model and is selling it. That does not make them false. It does mean outsiders will have to measure the model themselves before anyone can call the claims settled.

The price of the light element

For developers, Google is charging $1.50 per million input tokens and $7.50 per million output tokens. The company says the price stays the same across the whole one-million-token context window. In a post for developers, Google pitches Hydrogen for coding and debugging inside code repositories, interactive web development and work that mixes text, images and other media. It can run alone or alongside Carbon and Argon.

That last option shows what Hydrogen is really for. The Antigravity team introduced a mode called Orchestrator, which pairs Hydrogen with Google’s strongest models, including Carbon, on long tasks. In Google’s Terminal-Bench 4.0 test of that mode, the company says the pairing came close to Carbon’s task success while using 24 percent less quota on average and finishing 19 percent faster.

The heavy elements still get the first look and the paying customers. Hydrogen, the lightest, has been told to start working, and to do it everywhere at once.