Google officially launches Gemini 4 Hydrogen model to enhance search service intelligence

·by Henderson·AI
Google officially launches Gemini 4 Hydrogen model to enhance search service intelligence

On November 12, local time, Google officially deployed the next-generation model Gemini 4 Hydrogen to its search service, making it available to users worldwide. Google Search Vice President Robby Stein described it on social media as "the largest model upgrade for Google Search to date." At the same time, Gemini 4 Hydrogen replaces the previous Gemini 3.8 Flash as the default model for AI Mode and begins to handle the most complex queries in AI Overviews.

Google positions Carbon and Argon as flagship models for handling deep, long-range reasoning tasks, while Hydrogen aims to achieve fast execution in a wide range of everyday tasks, particularly in coding, tool invocation, and multi-step intelligent workflows. Google's logic is that not every question requires the most powerful model; a faster, more cost-effective, and sufficiently intelligent model can be a better solution in many scenarios. The speed improvement of Gemini 4 Hydrogen is significant, with Google stating that Hydrogen completes tasks 2.5 times faster than the previous fast model, Gemini 3.8 Flash.

Gemini 4 Hydrogen enhances search intelligence and flexibility

However, beyond speed, its intelligence level has not been compromised: at the default high reasoning level of the Artificial Analysis Intelligence Index, Hydrogen scored 51, even slightly higher than the higher-positioned Gemini 4 Argon (50); at the newly added "xhigh" highest reasoning level, the score can reach 54. The most direct change brought by Gemini 4 Hydrogen is a significant increase in the tolerance for "imperfect queries" in search.

Google states that the new model can handle more nuanced and complex queries—including compound questions with multiple parts, highly specific requests, and difficult-to-understand spelling errors.

Users no longer need to trim their thoughts into keywords but can pose complete questions in a natural flow of thought, with the model breaking down the various parts of the question and responding to each one. This capability covers all languages supported by AI Mode. Query Fan-out is a mechanism in search that breaks down a single user query into multiple sub-queries in the background, retrieving relevant information in parallel. Hydrogen's advanced reasoning capabilities enhance this technology, allowing it to execute simultaneous sub-topic queries more accurately. Google's description is that the result is "connecting you to a broader range of useful links than ever before," including websites that users might never think to visit themselves.

With the addition of Google Search Hydrogen, the response formats in AI Mode have become richer and more structured. Users will see more images, structured answers, comparison tables, and customized tools as visual responses to queries. A specific case mentioned in a report by The Verge describes how when a user asks, "How does a phone determine which lane a vehicle is in on the highway without cellular signal?" the AI generates an interactive dynamic driving simulation instead of a block of explanatory text. The goal of the responses has shifted from "reading a paragraph" to "grasping the structure at a glance."

Beyond search, Gemini 4 Hydrogen has simultaneously entered all levels of the Gemini application, Google AI Studio, Gemini API, and the enterprise-focused Gemini for Business. Free users can also use Hydrogen in the Gemini application through automatic model selection. In terms of pricing, Hydrogen adopts a fixed price within a full 1 million Token context window: US$1.50 per million input Tokens (approximately HK$12), and US$7.50 per million output Tokens (approximately HK$58).

ItemSpecification
ModelGemini 4 Hydrogen
Speed Improvement2.5 times faster than Gemini 3.8 Flash
High Reasoning Level Score51
xhigh Reasoning Level Score54
Input Token PriceUS$1.50 (approximately HK$11)
Output Token PriceUS$7.50 (approximately HK$58)

H
About the author
Henderson