Google has officially rolled out Gemini 4 Hydrogen, a new artificial intelligence model built to enhance the speed and efficiency of everyday AI tasks, coding, and agentic workflows. According to the company, the model completes tasks 2.5 times faster than its predecessor, the Gemini 3.8 Flash.
A core feature of Gemini 4 Hydrogen is its four variable thinking levels—low, medium, high, and extra-high. This system allows users to select the reasoning depth required for a specific task, meaning they only wait and pay for the exact depth of processing needed.
In terms of pricing, Google has set the cost at $1.50 per million input tokens and $7.50 per million output tokens across its one-million-token context window. Internal evaluations by Google indicate that Hydrogen more than tripled its score on the Terminal-Bench 4.1 coding benchmark compared to Gemini 3.8 Flash, achieving 57 per cent against the older model's 18.2 per cent. Furthermore, the company reported that the new model reduced fabricated claims by more than five times.
Alongside Hydrogen, Google is introducing Orchestrator in Antigravity, a feature designed to dynamically switch between AI models during a session depending on the requirements of the task. On Terminal-Bench 4.0, Orchestrator achieved task success rates comparable to Gemini 4 Carbon while using 24 per cent less quota on average and completing tasks 19 per cent faster.
The company stated that Hydrogen is capable of working alongside other models in a collaborative setup, where one model handles coordination while others manage specific tasks such as coding, testing, and verification. The model is also engineered to support research, data analysis, and document-heavy tasks involving spreadsheets, charts, audio, and video.
This release follows earlier rollouts within the Gemini 4 family. Google previously introduced Gemini 4 Argon for complex applications in software engineering, finance, legal services, and scientific research, alongside Gemini 4 Carbon, which targets demanding, long-running workflows like cybersecurity and scientific reasoning.
Additionally, Google noted that Gemini 4 Hydrogen includes notable safety upgrades. The model reportedly reduces excessive actions in agentic tasks by up to 81 per cent and is up to twice as unlikely to bypass permission boundaries when encountering blocked or impossible tasks.
"The introduction of tiered reasoning levels and improved processing speeds reflects a practical shift in the AI market toward operational efficiency and cost control. For businesses and startups integrating AI into their workflows, having granular control over token consumption and reasoning depth allows for better resource management and predictable scaling." — Dr. Shishir Gupta, Founder & CEO, StartupLanes
Recent StartupLanes Articles
Browse through our 30 latest publications on venture capital, startups, and angel investing.