Google releases Gemini 4 Hydrogen for coding, Search and the Gemini app

Google is bringing its newest AI model to everyday work, promising faster execution alongside strong reasoning. Gemini 4 Hydrogen began rolling out on November 12, 2026, across coding tools, business products, the Gemini app and Search. The performance case rests on company-published benchmarks, while access and API pricing are now concrete.

A fast worker, with adjustable reasoning

Google draws a division of labor within its Gemini 4 family. Carbon and Argon are positioned for difficult problems requiring extended reasoning; Hydrogen targets rapid execution. It can work alone on repository coding, debugging and interactive web development, or handle writing code, running tests and checking changes alongside those models.

The API costs $1.50 per million input tokens and $7.50 per million output tokens, with flat rates across the full 1-million-token context window. Google also lists up to 1 million output tokens, positioning Hydrogen to turn PDFs, spreadsheets, charts, audio and recorded video into reports and structured deliverables.

Hydrogen offers four thinking levels: low, medium, high and xhigh. High is the default. Google presents the settings as a way to balance response time, reasoning depth and cost. It reports an Artificial Analysis Intelligence Index score of 51 at high and 54 at xhigh, compared with 50 for Argon.

For teams moving from Gemini 3.8 Flash, Google claims Hydrogen finishes tasks 2.5 times faster at its default setting and cuts fabricated claims by more than fivefold. Its published Terminal-Bench 4.1 comparison also shows a substantial increase in the score.

Sharing a session with Carbon

The release also introduces Orchestrator in Google Antigravity. Instead of keeping one model on a task from beginning to end, it dynamically switches between models such as Carbon and Hydrogen within a session. Google reports near-Carbon task success on Terminal-Bench 4.0, with 24% less average quota consumption and 19% faster completion.

Google describes another configuration for parallel work: Carbon coordinates, while Hydrogen powers worker and verifier agents. Those agents can build and test changes simultaneously in isolated workspaces. That gives Hydrogen a defined supporting role as well as a standalone one, rather than making it a replacement for every Carbon assignment.

For agents that take actions, Google also claims safety improvements over 3.8 Flash. It reports a twelvefold reduction in misaligned actions and task shortcutting across tool-use and coding evaluations. On its internal agent honesty evaluation, Hydrogen achieved 96.7%; against indirect prompt injection, Google reports more than 2.3 times greater resilience on the Gray Swan IPI benchmark.

Where Hydrogen is rolling out

• Developers: Access is through the Gemini API in AI Studio and the paid tier in Android Studio. Google AI Pro and Ultra subscribers can also use Hydrogen in Antigravity.
• Businesses: Hydrogen is rolling out in Gemini for business through the API or model selector, and in Gemini in Google Sheets for eligible business subscriptions.
• Consumers: The rollout includes the Gemini app’s model selector and Google Search for all users globally. Sheets access starts with Google AI Pro and Ultra subscribers.

In the Gemini app, Google says Hydrogen follows tone, formatting and other instructions more closely, and more often selects the right tools. Its task workflows can draw context from uploads, Gmail, Google Drive and connected services such as Airtable and Dropbox. The promise is less back-and-forth when producing drafts, research and other everyday work.

Search gets a different application of the model: better handling of multipart questions, typos and specific requests, alongside more visual responses. Google also says Hydrogen upgrades query fan-out—the technique Search uses to pursue related queries—to surface a broader range of relevant links.