Google Launches Gemini 4 Hydrogen: A Faster, Cheaper AI Model for Everyday Coding and Agents

Google has launched Gemini 4 Hydrogen, a new AI model built for speed. Announced on November 12, 2026, it is the fast, affordable member of the Gemini 4 family, aimed at everyday coding work and AI agents that carry out tasks with tools. While Gemini 4 Carbon and Argon handle the deepest reasoning jobs, Hydrogen is the model Google wants developers and everyday users to run all day long.

The launch matters because it brings Gemini 4 to everyone, not just paying subscribers. Hydrogen is rolling out across Google AI Studio, the Gemini API, the Gemini app, Google Search and Gemini for business, with paid Google AI Pro and Ultra subscribers also getting it in Antigravity. Google says it is a major upgrade over Gemini 3.8 Flash, the fast model it replaces.

What Gemini 4 Hydrogen is built for

Hydrogen is designed for rapid execution. Google describes it as the workhorse of the Gemini 4 family: quick at everyday coding, debugging, issue resolution and codebase questions, and steady when an AI agent needs to call tools, read documents or search the web. It can take in text, images, audio and video, and it works across a one million token context window.

The model is meant to stand on its own or pair with its bigger siblings. In multi-model setups, Carbon or Argon can plan a complex task while Hydrogen carries out the steps, acting as what Google calls the high-efficiency execution engine. That split is already live in Antigravity, Google’s coding environment, where a new Orchestrator feature switches between Carbon and Hydrogen inside one session.

How fast it is, according to Google

Google’s headline claim is speed. At its default high reasoning setting, Hydrogen finishes tasks 2.5 times faster than Gemini 3.8 Flash. On Terminal-Bench 4.1, a benchmark that tests real coding work in a terminal, Google reports a score of 57 percent, against 18.2 percent for Gemini 3.8 Flash. The company also says Hydrogen cuts fabricated claims by more than five times compared with Flash, and adapts cleanly when a tool call fails or a search comes back empty.

On the independent Artificial Analysis Intelligence Index, Hydrogen scores 51 at its default high setting, just ahead of Argon at 50, and reaches 54 at its highest thinking level. Those numbers are self-reported by Google for its own tests, and independent evaluators have noted that moving above the default setting raises the score while also increasing task time and cost.

Four thinking levels and flat pricing

Hydrogen introduces four thinking levels: low, medium, high and a new xhigh tier. High is the default, and the idea is simple: you pick how much reasoning a task deserves, and you only wait and pay for that depth. A quick code fix can run at low, while a trickier debugging job can step up to xhigh.

Pricing is flat across the whole one million token window, with no long-context surcharge. Hydrogen costs 1.50 dollars per million input tokens and 7.50 dollars per million output tokens. Google positions that as cheap enough to run agents all day, and the model is available on free tiers of AI Studio and the Gemini API, with rate limits, as well as on paid tiers.

Where you can try it today

Hydrogen is rolling out globally across Google’s developer, enterprise and consumer surfaces. Developers can call it through the Gemini API with the model ID gemini-4-hydrogen, or try it in AI Studio. Everyday users will find it in the Gemini app and in Google Search, where it is picked by automatic model selection. Paid Google AI Pro and Ultra subscribers can also use it in Antigravity, and Gemini in Sheets access is starting on those plans.

In Antigravity, the new Orchestrator feature lets Carbon coordinate parallel agents while Hydrogen powers worker and verifier agents in separate workspaces. According to Google, that setup approached Carbon’s task success on Terminal-Bench 4.0 with 24 percent less quota and 19 percent faster average completion.

The fine print worth knowing

Hydrogen is not the smartest model Google sells, and Google is open about that. Carbon remains the strongest model in the family, and Argon is built for long, drawn-out tasks with lower token use. Hydrogen trades some depth for speed and price, which is the right call for everyday work but not for every job.

There are also some early caveats. Google DeepMind’s model card says Hydrogen is adapted from Carbon, with a knowledge cutoff of March 2026, and reports more unjustified refusals than Flash, particularly on prompts that include images or other non-text input. Google says the full Safety and Security Report will be available in the coming days. As with any new model, the benchmark numbers come from Google’s own testing, so real-world results will vary by task.

Even so, the launch fills a clear gap. Gemini 4 Carbon arrived first for heavy reasoning, and Argon followed for long-horizon work. Hydrogen is the everyday tier, and by putting it in Search and the Gemini app for free, Google is betting that most people will meet Gemini 4 through this fast model first.

Frequently Asked Questions

What is Gemini 4 Hydrogen?

Gemini 4 Hydrogen is Google’s fast, lower-cost AI model for everyday coding and agentic workflows, announced on November 12, 2026. It sits below Gemini 4 Carbon and Argon in the Gemini 4 family, trading some depth for speed and price.

How much does Gemini 4 Hydrogen cost?

Hydrogen costs 1.50 dollars per million input tokens and 7.50 dollars per million output tokens, flat across its one million token context window. It is also available on free tiers of Google AI Studio and the Gemini API, with rate limits, and in the Gemini app and Google Search.

Where can I use Gemini 4 Hydrogen?

You can use it through the Gemini API and Google AI Studio, in the Gemini app and Google Search, in Gemini for business, and in Antigravity if you have a paid Google AI Pro or Ultra subscription.