Google Rolls Out Gemini 4 Hydrogen Across Search, Its App And API

Google announced Gemini 4 Hydrogen on 12th November 2026, extending its Gemini 4 models to Search, the Gemini app and developer services. In its company blog post, Google describes Hydrogen as a model for everyday coding and workflows in which software uses tools to complete tasks. Developers can access it through the Gemini application programming interface (API), which lets applications send requests to the model.

The company says the rollout covers all tiers of Google AI Studio and Gemini for business. Its developer announcement also lists Android Studio and access through Google Antigravity for paid Google AI Pro and Ultra subscribers. Free users of the Gemini app receive Hydrogen through automatic model selection when the system considers it appropriate for their prompt, rather than choosing it directly.

Google is also introducing Orchestrator in Antigravity to switch between models such as Carbon and Hydrogen within a session. Google says Carbon can coordinate the work while Hydrogen can build and test changes in separate workspaces.

For Search, Google says Hydrogen is available globally to all users. Its Search post promises better handling of questions with several parts and more visual answers, including comparison tables and images. Google also says it has improved how Search runs simultaneous queries across related topics to find relevant links. These are Google's descriptions of the update, rather than independently measured improvements in search quality.

Reasoning settings change cost and speed

Google lists a flat price across Hydrogen's entire context window of one million tokens, the capacity for material processed in a request. Input costs $1.50 per million tokens and output costs $7.50 per million. Tokens are the units used to measure model input and output. Hydrogen offers low, medium, high and xhigh reasoning settings, with high as the default. Google says these settings adjust how deeply the model works through a task, with trade-offs in answer quality, waiting time and cost.

The Artificial Analysis results for high reasoning, read on 13th November at 03:32 UTC, give Hydrogen an Intelligence Index score of 51. The index combines nine evaluations. Artificial Analysis reports 8.3 minutes per index task and an average cost of $1.87 per task at that setting.

Its medium setting results show a score of 41, with 3.7 minutes and $0.80 per task. At the low setting, the score falls to 27, while time falls to 1.3 minutes and cost to $0.29 per task. These are benchmark results, rather than estimates for every application. They show why an unchanged price per token does not mean an unchanged bill for completing a task.