Google said on Nov. 12 it launched Gemini 4 Hydrogen, a new AI model that outscored the larger Argon on benchmarks just four weeks after Argon's release.
Hydrogen scored 51 points on the Artificial Analysis Intelligence Index, against Argon's 50, and reached 54 at its maximum xhigh reasoning setting, Google said in a blog post. The company said the model runs 2.5 times faster than the previous Gemini 3.8 Flash and scored 57% on Terminal-Bench 4.1, versus 18.2% for its predecessor.
Fabricated answers dropped by a factor of five, the company said. Developers pay $1.50 for every million tokens Hydrogen reads and $7.50 for every million it generates, within a context window of one million tokens.
Koray Kavukcuoglu, senior vice president at Google DeepMind and the company's chief AI architect, said Hydrogen “is built to stand on its own as a model for everyday work,” adding that it can also work alongside Carbon and Argon, serving as the fast layer when several models combine on one task.
The model is already live worldwide in Google Search, the Gemini app and the Antigravity developer tool, according to the company's blog.
Google Search's vice president of product, Robby Stein, said the tool “more precisely follows what you want and gives you relevant responses that address each part.”
Independent lab Artificial Analysis found Hydrogen fabricates answers only 8% of the time, though it lags behind Claude Sonnet 5.5 and GPT-6.1 Sol on terminal-based tasks.
Hydrogen is the third Gemini 4 model in six weeks: Google introduced the flagship Argon on Oct. 1 after six months of delays, then released the security-focused Carbon on Oct. 28.


