Google releases Gemini 4 Hydrogen, a lower-cost model

Google says Hydrogen scored 51 against Argon's 50 on the Artificial Analysis Intelligence Index at its default high thinking level. A new fourth thinking level, xhigh, raises Hydrogen's score to 54. Unlike Argon, which first went to a group of vetted cybersecurity defenders, Hydrogen skipped that stage and opened to everyone on launch day.

Google also added Orchestrator to its Antigravity coding tool, which switches between Hydrogen and Gemini 4 Carbon during a session. In Google's Terminal-Bench 4.0 runs, Orchestrator came close to Carbon's success rate while using an average of 24% less quota. Compared with Gemini 3.8 Flash, Google says Hydrogen finished tasks 2.5 times faster at the default setting and scored 57% on Terminal-Bench 4.1, against 18.2% for the older model.

Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens across its full 1 million-token context window. Argon is priced at $2 and $10 until its introductory pricing ends on Nov. 30, after which it rises to $4 and $20.

Developers building agents get a cheaper Google model that they can call through the Gemini API without waiting for a paid-only rollout. Hydrogen's per-token prices are lower than Argon's post-introductory rates, which matters for agents that make many calls per task.

Google's benchmark figures come from its own runs and are not independently verified in this report, so teams should test Hydrogen on their own workloads before switching. The xhigh thinking level may also change latency and cost per request, so it is worth measuring.