Google introduces Gemini 4 Hydrogen for everyday AI tasks

Google is rolling out Gemini 4 Hydrogen, an AI model designed for everyday tasks, coding and work that requires several steps. It is coming to the Gemini app, Google Search and developer and business products worldwide.

Google positions Hydrogen as a faster, more efficient option alongside its Carbon and Argon models, which it recommends for more complex reasoning. Free Gemini app users can access Hydrogen through automatic model selection when Google’s system considers it suitable for their request. Paid users can select it in the model menu.

For content work, Google says the model is better at following instructions about tone, format and other requirements. It can also draw on material from uploads and connected services, including Gmail, Google Drive, Airtable and Dropbox, when relevant connections are available.

Hydrogen can work with PDFs, spreadsheets, charts, audio and video. Google says it can turn such material into reports with citations and other structured documents. The company also shows examples of the model producing a slide presentation from a video and a design template, and helping create financial forecasts in Google Sheets.

Faster performance and more control

Google reports that Hydrogen completes tasks 2.5 times faster than its predecessor, Gemini 3.8 Flash, at its default setting. It also says the model makes more than five times fewer fabricated claims. These are Google’s claims, and results in everyday use may vary.

The company offers four thinking levels: low, medium, high and xhigh. High is the default. The settings let developers balance the amount of reasoning against response time and cost.

For developers, Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens. Tokens are the units models use to process and generate information. The model supports a context window of one million tokens and can generate up to one million output tokens, according to Google.

Google is also introducing Orchestrator in its Antigravity development product. It switches between models during a session, allowing Carbon to handle demanding decisions and Hydrogen to take on routine execution. On one coding benchmark, Google says this approach achieved results close to Carbon alone while using 24% less quota and finishing 19% faster on average.

Search, availability and safety

In Search, Google says Hydrogen improves its ability to understand complicated questions and provide more visual answers and a wider range of links. According to 9to5Google, Hydrogen is the default model for AI Mode and is also used for the hardest questions in AI Overviews.

The rollout also includes Gemini for business, the Gemini API and Google AI Studio. Access in Antigravity requires a paid Google AI Pro or Ultra subscription. In Google Sheets, availability depends on the subscription.

Google says it has improved the model’s safety when it uses tools or acts on a user’s behalf. In its own tests, the company reports fewer instances of the model exceeding its authority or taking shortcuts. It also says Hydrogen is more resistant than 3.8 Flash to instructions hidden in outside material that attempt to redirect an AI agent.

Sources

• Introducing Gemini 4 Hydrogen – Google
• Google announces Gemini 4 Hydrogen with ‘significant leaps’ in speed, intelligence – 9to5Google