Google newly revealed Gemini 4 Hydrogen, a new artificial intelligence model designed to accelerate daily tasks and advanced uses. It comes to the Google Gemini app, API, and search engine with an important promise: to make the capabilities of Gemini 4 accessible to as many people as possible.
Hydrogen completes the Gemini 4 range, already made up of two models, Argon and Carbon. While the latter favor deep reasoning and complex tasks, Hydrogen focuses on speed of execution and efficiency. It thus resumes the positioning historically occupied by Flash models, without retaining this name in the new generation.
Gemini 4 Hydrogen focuses on speed and AI agents
Google specifically targets its new software development model, in-depth research and agent workflows. These allow an AI to divide a task into several steps, use external tools and execute different actions without constantly prompting the user.
Gemini 4 Hydrogen has a million token popup. The template supports prompts for text, images, audio, videos, and PDF documents. Google also announces four levels of reasoning: Low, Medium, High and Xhigh. In this way, the user can adapt the resources dedicated to a task depending on its complexity.
The benchmarks reported by Google are ambitious. The company claims that Gemini 4 Hydrogen completes certain tasks 2.5 times faster than Gemini 3.8 Flash, while scoring 57% in Terminal-Bench 4.1, compared to 18.2% for its predecessor. Google also announces a more than five-fold reduction in complaints filed for the model. Furthermore, benchmarks show that Gemini 4 Hydrogen still performs better than OpenAI’s GPT-6.1 Terra and almost always wants to become Anthropic’s Claude Sonnet 5.5.
Artificial Analysis also gives you a score of 51 on your Intelligence Index at its High setting. This independent index is based on nine assessments covering software development, knowledge and agency capabilities. The result places the Gemini 4 Hydrogen above the average of comparable models in terms of price, set at 41 points.
Google search engine also benefits from Gemini 4 Hydrogen
The other important novelty refers to the Google search engine. Gemini 4 Hydrogen should improve query understanding, especially when users make long requests with multiple questions or typos. Google also promises more visual answers and a broader selection of relevant links.
To do this, the company is significantly improving its “query fan-out” technique, which consists of breaking down a search into several complementary queries to explore different avenues. The goal is to provide more relevant answers and show pages that would have otherwise gone unnoticed. Gemini 4 Hydrogen is rolling out in Google AI mode globally.
In the Gemini app, subscribers can directly select Hydrogen from the model menu. Free users will also have access to it through automatic selection when Google deems this model suitable for their request. Developers can use it in Google AI Studio and through the Gemini API. It is also offered to businesses and Google AI Pro and Ultra subscribers in Google Antigravity.
On the developer side, Google announces a price of $1.50 per million input tokens and $7.50 per million output tokens. This price positioning should allow Gemini 4 Hydrogen to be used in large volumes without incurring the cost of the most powerful models in the range.
