Google rolls out faster Gemini 4 Hydrogen for coding, agents

Google announced the release of Gemini 4 Hydrogen, a more affordable model in its Gemini 4 artificial intelligence series. According to an independent intelligence ranking, Hydrogen outperformed the existing Gemini 4 Argon model by achieving a score of 51 compared to Argon's 50 on the Artificial Analysis Intelligence Index. The advanced xhigh thinking level further boosts Hydrogen's performance, scoring 54.

Unlike its predecessor, Hydrogen was made available immediately without the Fairwind Program's cybersecurity scrutiny. Upon release, anyone, including those on the free Gemini app, could use the model. Google's DeepMind head Koray Kavukcuoglu highlighted Hydrogen's design for swift execution. Google's Antigravity coding tool integrated Orchestrator, which now decides between Hydrogen and Gemini 4 Carbon during sessions.

Google Terminal-Bench 4.0 tests revealed Hydrogen completing tasks 2.5 times faster than Argon at default settings, with a notable 57% score on Terminal-Bench 4.1 command line tests. Early adopters like former OpenAI scientist Andrew Carr and Discovery Loop Inc., co-founded by former Google Chief Scientist Jeff Dean, praised Hydrogen's problem-solving capabilities.

Google plans to introduce Hydrogen in the Gemini API, AI Studio, and Search services, with paid subscribers enjoying the model in Antigravity. Pricing starts at $1.50 per million input tokens and $7.50 per million output tokens, with Argon's previous rates set to increase upon expiration in November.