Google has announced the launch of Gemini 4 Hydrogen, a mid-tier language model developed to deliver agility and efficiency in everyday tasks and agentic workflows. The new artificial intelligence joins the family of models based on elements of the periodic table, positioned alongside Gemini 4 Carbon and Gemini 4 Argon. Unlike ultra-high-performance options, Hydrogen was designed to serve a broader scale of users and demands.
According to Google, the model represents a significant advancement in speed and intelligence compared to its predecessor, Gemini 3.8 Flash. In a high-effort configuration, the new version executes tasks 2.5 times faster and achieved more than triple the score on Terminal-Bench 4.1, reaching 57% compared to 18.2% for its predecessor. In comparative tests released by the company against mid-tier competitors such as Claude Sonnet 5.5 and GPT-6.1 Terra, Google's model outperformed rivals in most evaluations, although it lagged behind Anthropic's model in agentic programming, computer use, and math and science tests.
To accompany the launch, the company introduced the Orchestrator tool within the Google Antigravity ecosystem, capable of dynamically switching between Carbon and Hydrogen models depending on task complexity. In Terminal-Bench 4.0 tests, Orchestrator matched the performance of Gemini 4 Carbon with a 24% reduction in usage quota consumption and 19% faster task completion.
Gemini 4 Hydrogen is distributed across different channels in the Google ecosystem. Developers have access via the Gemini API, Android Studio, and AI Studio, while subscribers to the Google AI Pro and AI Ultra plans can use it on Google Antigravity. For free plan users and the Google AI Plus version, the model is accessible through the Gemini app and Search. The company set the cost at $1.50 per million input tokens and $7.50 per million output tokens. Starting Friday, November 13, 2026, AI Plus plan subscribers will no longer have the option to select Gemini Pro category models.