Google rolls out Gemini 4 Hydrogen with 2.5x speed boost, four reasoning levels

Google has rolled out Gemini 4 Hydrogen, a new artificial intelligence model that completes tasks 2.5 times faster than its predecessor, Gemini 3.8 Flash, as the company looks to improve the speed and efficiency of everyday AI tasks, coding and agentic workflows.

The model is being made available across Google AI Studio, the Gemini API, Gemini for business, the Gemini app and Google Search. Paid Google AI Pro and Ultra subscribers can also access it through Google Antigravity, according to the company.

In a post on X, Google CEO Sundar Pichai said the model was built for speed and efficiency across everyday tasks, coding and agentic workflows.


"Compared to 3.8 Flash, it finishes tasks 2.5x faster. Also has 4 thinking levels, so you only wait and pay for the reasoning depth a task requires," he said.

Gemini 4 Hydrogen offers four thinking levels -- low, medium, high and extra-high -- allowing users to choose the depth of reasoning based on their requirements. Google said the model is priced at USD 1.50 per million input tokens and USD 7.50 per million output tokens across its one-million-token context window.

According to Google's evaluations, Hydrogen more than tripled its score on the Terminal-Bench 4.1 coding benchmark compared with Gemini 3.8 Flash, scoring 57 per cent against 18.2 per cent. It also reduced fabricated claims by more than five times.

Google is also introducing Orchestrator in Antigravity, a feature that dynamically switches between AI models during a session depending on the task. On Terminal-Bench 4.0, Orchestrator achieved task success rates close to those of Gemini 4 Carbon, while consuming 24 per cent less quota on average and completing tasks 19 per cent faster.

The company said Hydrogen can work alongside other models, with one handling coordination and others carrying out tasks such as coding, testing and verification. It is also designed for research, data analysis and tasks involving documents, spreadsheets, charts, audio and video.

The launch follows Google's rollout of other models in the Gemini 4 family. Gemini 4 Argon was introduced for complex work across software engineering, finance, legal services and scientific research, while Gemini 4 Carbon was positioned for more demanding, long-running workflows, including cybersecurity and scientific reasoning.

Google further said Hydrogen also includes improvements in safety, reducing excessive actions in agentic tasks by up to 81 per cent and making the model up to twice as unlikely to bypass permission boundaries when faced with blocked or impossible tasks.