← All news·2026-11-12·3 min read

Google Launches Gemini 4 Hydrogen for Agentic Tasks

Google has released Gemini 4 Hydrogen, an AI model built for agentic work. It is cheaper and faster than the company's flagship models, and it comes with better accuracy and stronger protection against prompt injections.

aigoogle

Google released Gemini 4 Hydrogen, a new artificial intelligence model built for agentic work — tasks where the system takes several steps on its own, like running code or checking its own results. It sits below the company's top-tier models, called Carbon and Argon. Hydrogen is already built into Google Search and other Google products. Google says the model nearly tripled its score on Terminal-Bench 4.1, a test of multi-step coding tasks, from 18.2% to 57%. The company also says it runs 2.5 times faster than the previous model, Gemini 3.8 Flash, and makes up facts less than a fifth as often.

For people building software with these tools, a cheaper, faster model that still handles multi-step tasks well lowers the cost of using artificial intelligence as a working assistant, not just a chat window. Google also says the model strays outside its assigned task about 81% less often than before — a problem that matters for anyone wary of letting these systems act on their own. For ordinary users, the practical change is that these agentic abilities move into everyday tools like Search, rather than staying locked inside a separate product few people open.

Source: blog.google

Free course

Stop reading about AI — start building with it

The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.

Start free →
EAEvgenii Arsentev

Author

Evgenii Arsentev

PhD · Founder & CEO, Dad4Kids