Google unveils new Gemini 4 Hydrogen model, aimed at speed and efficiency

On this week's Carbon availability, Google has introduced Gemini 4 Hydrogen, replacing the "Flash" branding. This new model is described as bringing "cutting-edge intelligence to everyone" and is touted as a "major standalone upgrade" with significant improvements in speed and intelligence.

In comparison to Gemini 3.8 Flash, launched in September, Hydrogen offers a native 1 million-token input and output capability. It surpasses Argon on the Artificial Analysis Intelligence Index, with scores of 51 and 54 at the default high and highest thinking levels respectively, while often completing tasks faster than models within its class.

The Gemini app benefits from Hydrogen's enhanced ability to understand and adhere to user-provided tone, format, and rules. It generates clear tables and visual breakdowns, enabling users to receive email drafts, study guides, rewrites, and other documents ready for immediate use. Additionally, Hydrogen frequently selects the correct tools and skills for specific tasks.

When users select the 4 Hydrogen model in the model picker, it is described as providing "Everyday help." Free users in the Gemini app have access to Hydrogen via the "[Auto] model selection" when it's deemed the most suitable for the given prompt.

Google claims that Hydrogen significantly enhances Search (AI Mode) by better understanding user intent. Users can ask questions in a natural flow of thought, even with multiple parts, hard-to-understand typos, or highly specific requests, and Search will intelligently interpret the query. The responses will be more visual and include a wider range of relevant links.

Hydrogen is accessible globally in Google Search for all users. In Google Antigravity, it features an "Extra" thinking tier alongside Low, Medium, and High. Additionally, there's Orchestrator, which handles routine steps offloading to Hydrogen while maintaining the same Carbon-level quality with fewer tokens used.

Orchestrator is introduced to facilitate seamless multi-model collaboration in Google Antigravity. It dynamically switches between models like Carbon and Hydrogen within a single session. Carbon acts as the coordinator in workflows that require parallel agents, while Hydrogen powers worker and verifier agents to build and test changes simultaneously across isolated workspaces, thereby speeding up complex project completion without escalating token costs.

For developers, Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens. Hydrogen is available in Antigravity for paid Google AI Pro and Ultra subscribers.