Google Launches Gemini 4 Hydrogen AI Model for Coding and AI Agents

Google has introduced Gemini 4 Hydrogen, a new model aimed at developers and businesses that want quick results from everyday coding and automated “agent” tasks. The announcement came on November 12 from Koray Kavukcuoglu, who leads AI architecture at Google. Hydrogen sits beneath the company’s two heavier-duty models, Gemini 4 Carbon and Gemini 4 Argon, which are meant for slow, deep, long-running reasoning.

Speed without giving up much intelligence

Google says Hydrogen scores 51 on the Artificial Analysis Intelligence Index at its default setting, just above Argon’s 50, and 54 when pushed to its new top thinking level. Compared with the earlier Gemini 3.8 Flash, Google reports that Hydrogen finishes tasks about two and a half times faster and more than triples its score on the Terminal-Bench 4.1 coding test, rising from 18.2% to 57%.

Google also cites a long list of other results, including 58.8% on HLE-Diamond, first place...

Copyright of this story solely belongs to www.itvoice.in. To see the full text click HERE