Google announced a new AI model, Gemini 4 Hydrogen, on November 12, 2026 U.S. time (November 13 JST). The company describes the fast, lower-cost model as "our daily driver for agentic workflows," and it began rolling out worldwide the same day across Google's developer, enterprise, and consumer services. The announcement was published on the official blog under the name of Koray Kavukcuoglu, SVP, Google DeepMind and Chief AI Architect, Google.

Hydrogen Is the Gemini 4 Model Built for Speed and Efficiency

The Gemini 4 series already includes two top-tier models, Gemini 4 Carbon and Gemini 4 Argon, which handle deep, long-horizon reasoning. Hydrogen sits below them and is meant to execute everyday coding and agentic work quickly.

According to Google's published figures, Hydrogen at its default thinking level completes tasks 2.5 times faster than the earlier Gemini 3.8 Flash. On Terminal-Bench 4.1, which measures how well a model operates in a terminal, it scored 57%, more than three times Gemini 3.8 Flash's 18.2%. On the Artificial Analysis Intelligence Index, it scored 51 at the default thinking level, ahead of the higher-tier Argon at 50, and Google says it reaches 54 at the highest thinking level.

There are four thinking levels: low, medium, high, and xhigh. High is the default, and xhigh is new with this release. The context window is 1 million tokens.

API Pricing Is a Flat $1.50 for Input and $7.50 for Output

API pricing is $1.50 per million input tokens and $7.50 per million output tokens. Google says the price is "flat across its entire 1 million-token context window," so the unit price does not change with the length of the input.

The model is available in the following places.

  • Developers: the Gemini API in Google AI Studio, and Android Studio
  • Enterprises: Gemini for business and Google Antigravity
  • Consumers: the Gemini app and Google Search

In Google Antigravity, paid Google AI Pro or Ultra subscribers can use Hydrogen. Gemini in Google Sheets is also starting with Google AI Pro and Ultra subscribers.

Antigravity Gets Orchestrator, Which Switches Between Models

Google also announced that it is introducing Orchestrator in Google Antigravity, its development tool. Orchestrator moves work forward while switching between models such as Carbon and Hydrogen within a single session.

Google says that on Terminal-Bench 4.0, Orchestrator reached task success rates close to Carbon's while using 24% less quota on average and finishing 19% faster. For parallel work, Carbon can coordinate the overall job while Hydrogen runs the worker and verifier agents.

All Search Users Get Hydrogen, and Free Gemini App Users Get It Through Auto Selection

For Google Search, Robby Stein, VP of Product, Google Search, wrote in a separate post that Hydrogen is now available globally for all users. He calls it "the biggest model upgrade to Search yet." Search now reads the intent of multi-part questions and questions with typos more accurately, and it runs related searches simultaneously to surface a broader set of links. Visual responses that use images, comparison tables, and similar elements are also becoming more common.

In the Gemini app, Hydrogen can be chosen from the model dropdown. The free version uses automatic model selection, and Hydrogen is used when the app judges that it fits the prompt. According to a footnote in the announcement, most prompts in the free version are handled by Flash-Lite.

The announcement does not address availability in Japan specifically. Google describes the rollout as global, and says the Search improvements apply to all languages that AI Mode supports.

Our Take

What stands out here is the pricing structure more than the performance numbers. A flat rate that does not rise with long inputs makes it easier to estimate the cost of agentic work that loads large volumes of documents or code. Orchestrator, too, assumes a division of labor in which hard judgment calls go to a higher-tier model and high-volume work goes to a lighter one. In our view, Hydrogen is a model whose value comes less from its standalone intelligence than from being cheap and fast enough to run at scale.

That said, every benchmark figure cited here was published by Google itself. How well those results hold up in real work is something we would rather judge after third-party verification.

Summary

Google has announced Gemini 4 Hydrogen, a model that prioritizes speed and low cost, and has begun rolling it out worldwide across the API, the Gemini app, and Google Search. Pricing is a flat $1.50 per million input tokens and $7.50 per million output tokens. All Google Search users are covered, and the free version of the Gemini app also uses the model through automatic selection.