Google's Gemini 4 Hydrogen: AI Speed Redefined

Google has introduced Gemini 4 Hydrogen, its latest AI model engineered for exceptional speed and efficiency. This new model stems from Gemini 4 Carbon and is specifically crafted for rapid execution across various agentic workflows, making it ideal for tasks like writing code and managing daily work.

Hydrogen is designed to operate efficiently, either as a standalone AI for everyday tasks or as a high-performance engine when paired with Google's more advanced models like Gemini 4 Carbon and Argon. It allows users to adjust its "thinking level" across four tiers: low, medium, high, and a new "xhigh" setting, each offering different trade-offs in quality, latency, and cost.

The new AI model is rolling out globally starting today and is accessible to all users within Google Search. It is also available in the Gemini app's model selection menu. This advancement replaces Gemini 3.8 Flash, which previously powered Google Search's AI features. Hydrogen promises more visual responses, improved conversational understanding, and better adherence to user instructions, even with typos.

Google reports that Gemini 4 Hydrogen is designed to be more secure, being two times less likely to bypass permission boundaries compared to its predecessor. While the company continues to enhance its safety features, including resistance to jailbreaking, it acknowledges that, like all generative AI, Hydrogen may still produce hallucinations.