Google Launches Gemini 4 Hydrogen, Bringing Its Latest AI to Free Users

Google has launched Gemini 4 Hydrogen, bringing its latest generation of AI models to free users in the Gemini app and Google Search. The new model focuses on faster everyday assistance, coding, and tasks that require multiple tools.

The rollout is global. In the free Gemini app, access comes through automatic model selection when Hydrogen is appropriate for a prompt.

Everyday Help in Gemini and Search

In the Gemini app, Google promises better adherence to your requested tone, format, and instructions, along with clearer tables and visual breakdowns. Hydrogen can also draw on uploaded files and connected services such as Gmail, Google Drive, Airtable, and Dropbox to work through a task.

The Search upgrade is designed to handle more complicated questions, including multipart requests and typos, while producing more visual answers. It also improves the system that runs related searches in parallel, with the aim of returning a wider range of relevant links.

Google's demos show how the model can turn a prompt into a working interactive project. Geminisland creates structures in a shared 3D world that users can explore in first person.

Another example turns real-time lightning data into an interactive globe, mapping strikes to visual effects and music.

For document work, a Google Antigravity demo starts with a YouTube keynote and a template, generates a presentation in Google Slides, then checks the slides for overlapping text and layout problems.

Performance, Thinking Levels, and Pricing

Hydrogen joins Gemini 4 Carbon and Argon in Google's lineup. Google reports an Artificial Analysis Intelligence Index score of 51 at its default high thinking level, compared with Argon's 50, rising to 54 at the highest setting. Against Gemini 3.8 Flash, the company claims tasks finish 2.5 times faster and Terminal-Bench 4.1 performance rises from 18.2% to 57%.

Developers can choose low, medium, high, or xhigh thinking levels to adjust the amount of reasoning used for a task. The following charts show Google's comparisons of intelligence scores against completion time and cost.

The model supports a one-million-token input context and up to one million output tokens, with uses spanning PDFs, spreadsheets, charts, audio, and video. Google also reports a 76.1% score on AutomationBench-AA, which evaluates automation tasks.

In Google Antigravity, a new Orchestrator switches between models such as Carbon and Hydrogen within a session. In Google's Terminal-Bench 4.0 testing, that approach reached near-Carbon task success rates while using 24% less quota and finishing 19% faster on average.

API pricing is $1.50 per million input tokens and $7.50 per million output tokens, with the same rates across the full context window. Hydrogen is available through Google AI Studio, the Gemini API, and Gemini for business. Antigravity access requires Google AI Pro or Ultra; the rollout to Gemini in Google Sheets starts with those consumer plans and eligible business subscriptions.