AI3 min
Google releases Gemini 4 Hydrogen, a faster, cheaper AI model replacing 3.8 Flash
Google says Gemini 4 Hydrogen is built for quick everyday coding and agent tasks, while its bigger Carbon and Argon models handle hard reasoning. It costs $1.50 per million input tokens and $7.50 per million output tokens, at the same price across its 1-million-token context window. Google claims it more than triples Gemini 3.8 Flash's score on the Terminal-Bench 4.1 coding test (57% vs. 18.2%) and finishes tasks 2.5 times faster. It is now the default model in Search's AI Mode, and Google uses it for the hardest questions in AI Overviews. Google says it handles multi-part questions and typos better, and adds more images, tables and maps to answers. Independent tester Artificial Analysis rates it just below Claude Sonnet 5.5 and GPT-6.1 Sol. It costs a third as much as Sonnet but about 2.5 times as much as Sol. Google also added Orchestrator to its Antigravity coding tool, which switches between Carbon and Hydrogen within one session. Ars Technica notes that Hydrogen is not getting the slow, security-focused rollout that Carbon and Argon got.
Affects you: If you use Google Search's AI Mode or the Gemini app, you are already using it. Developers on 3.8 Flash get a cheaper upgrade path; the benchmark numbers are Google's own, except Artificial Analysis's ranking.