Google releases speed-focused Gemini 4 Hydrogen, with API pricing of $1.50 input and $7.50 output; also coming to the free Gemini app and Search

After Carbon and Argon, a model for everyday work

Google on Nov. 12 (Nov. 13 Japan time) released its AI model "Gemini 4 Hydrogen." While the higher-end "Gemini 4 Carbon" and "Gemini 4 Argon" take time to reason through hard problems, Google positions Hydrogen as a model that handles everyday tasks, coding and the work of AI agents that carry out tasks on their own quickly and efficiently. It is rolling out globally from the same day to developers, businesses and general users, including free users of the Gemini app and all users of Google Search.

Sundar Pichai, CEO of Google and Alphabet, announced it on X at 3:02:12 a.m. Nov. 13 Japan time. In the post, he said it finishes tasks 2.5 times faster than 3.8 Flash and offers a choice of four levels of thinking depth.

As an example from software development, Google described leaving design work and the analysis of difficult bugs to Carbon or Argon while Hydrogen handles writing code, running tests and verifying changes. Google said Hydrogen can be used on its own or together with Carbon and Argon.

The Hydrogen name had been posted on X as a codename since late October. On Nov. 5, leakers posted that a model was being tested under the name "Gemini 4 Flash" in the development environment "Google AI Studio" and elsewhere (earlier article). The official name Google announced is Gemini 4 Hydrogen, and the announcement does not use the Flash name.

API pricing is $1.50 input and $7.50 output

API pricing for developers calling the model from their own services is $1.50 for input and $7.50 for output per million tokens, the units of text the AI reads and writes. According to Google, the context window is 1 million tokens, and the same rate applies at any length within that range. The output price includes tokens for the thinking process before the model answers.

According to the Gemini API pricing page, the current Gemini 3.8 Flash costs $0.75 for input and $3.75 for output through Dec. 31, rising to $1.50 for input and $7.50 for output from Jan. 1, 2027. Hydrogen is priced the same as 3.8 Flash after that increase. The higher-end Gemini 4 Argon costs $2 for input and $10 for output through Nov. 30, and $4 for input and $20 for output from Dec. 1.

According to the pricing page, Hydrogen is also available on the Gemini API free tier. Argon is not available on the free tier. Discounts for repeatedly sent input (caching) are available only through the automatically applied method, at $0.075 per million tokens.

Google says it is 2.5 times faster than 3.8 Flash

According to Google, Hydrogen at its default setting finishes tasks 2.5 times faster than 3.8 Flash. On "Terminal-Bench 4.1," which measures programming work in a terminal, it scored 57% compared with 18.2% for 3.8 Flash, and Google said errors in which the model makes false claims fell to less than one-fifth. In early tests by partners, it ran 2 to 2.8 times faster than 3.8 Flash and used 12% to 23% fewer tokens, Google said.

Google said that on the Intelligence Index of U.S. evaluation firm Artificial Analysis, Hydrogen scored 51 at its default setting, above Argon's 50, and 54 at its deepest thinking setting. It also said Hydrogen matched or outperformed competing models of a similar class on key programming and business tasks. All of these figures come from Google's announcement, and this article does not cover third-party verification.

Four levels of thinking depth, plus automatic switching with Carbon

Hydrogen lets users choose how deeply it thinks before answering from four levels: "low," "medium," "high" and "xhigh." The default is high, and xhigh is newly added. Google said the design means users only wait and pay for the depth a task requires.

Google added "Orchestrator" to its development tool "Google Antigravity," which automatically switches between models such as Carbon and Hydrogen within a single task. According to Google, on Terminal-Bench 4.0 it kept a success rate close to Carbon's while using 24% less quota on average and finishing 19% faster. When multiple agents run in parallel, Carbon acts as the coordinator and Hydrogen builds and verifies changes.

Rolling out to developers, businesses and general users

Developers can use it on all tiers of Google AI Studio and the Gemini API. In the development tool Android Studio it is available on the paid tier, and in Antigravity it is available to paid Google AI Pro and Ultra subscribers.

Businesses can use it in "Gemini for business" through the API or the model selector, as well as in Antigravity and in Gemini in Google Sheets for eligible business subscriptions.

General users can choose it from the model selector in the Gemini app. For free users, Hydrogen is used when automatic model selection is on and it suits the prompt. In Google Search it is available to all users worldwide, and Google said it improves understanding of multi-part questions and surfaces a wider range of links. In Gemini in Google Sheets, it is rolling out starting with Google AI Pro and Ultra subscribers.

From rumor to official announcement

In the Gemini 4 line, Google announced Argon on Oct. 1 (Japan time) and Carbon on Oct. 29, and began general availability of Carbon on Nov. 11 (article on Carbon's general availability). Hydrogen was officially announced about two and a half weeks after the post about its codename.