Google launches Gemini 4 Hydrogen and rolls it into Search worldwide

Google has released Gemini 4 Hydrogen, a new addition to the Gemini 4 family built for fast execution rather than maximum reasoning depth, and has deployed it across nearly its entire AI portfolio — including Google Search, where it is live globally for all users.

What Google shipped

According to Google's announcement post, Hydrogen sits alongside Gemini 4 Carbon and Gemini 4 Argon, which the company describes as its frontier models for deep, long-horizon reasoning. Hydrogen targets the opposite end of the workload: everyday coding, tool use and agentic tasks where speed, cost and reliability matter more than peak intelligence.

The model is available in all tiers of Google AI Studio, the Gemini API, Gemini for business, the Gemini app, Google Search, and — for paid Google AI Pro and Ultra subscribers — in Google Antigravity.

Pricing and thinking levels

Google prices Hydrogen flat across its entire 1 million-token context window: $1.50 per million input tokens and $7.50 per million output tokens. It exposes four thinking levels — low, medium, high (the default) and a new xhigh tier — so users pay only for the reasoning depth a task requires.

Google says Hydrogen surpasses Argon on the Artificial Analysis Intelligence Index at its default high thinking level (51 versus 50), and reaches 54 at xhigh, while often finishing tasks faster than other models in its class.

The multi-model pitch

A central theme of the release is pairing Hydrogen with the frontier models rather than treating any single model as a universal tool. To automate that pairing, Google introduced Orchestrator in Google Antigravity, which switches between models such as Carbon and Hydrogen within a single session. On Terminal-Bench 4.0, Google reports that Orchestrator reached near-Carbon task success rates while consuming 24% less quota and completing tasks 19% faster on average. Carbon can act as a coordinator for parallel agents while Hydrogen powers worker and verifier agents, which Google argues lets teams ship complex projects faster without inflating token costs.

For teams moving off Gemini 3.8 Flash, Google claims Hydrogen finishes tasks 2.5 times faster at the high default setting, more than triples Terminal-Bench 4.1 scores (57% versus 18.2%), and cuts fabricated claims more than fivefold. The company also says the model adapts cleanly when a tool call fails or a search returns empty results.

The benchmark list is long. Google says Hydrogen outperforms similar-tier models on SWE-sweep, ranks fifth on arena.ai's live Agent Arena, leads its class on HLE-Diamond at 58.8%, places first overall on the Vals Tax Agent Bench (52.3%) and ties for first on the Vals Legal Research Bench (55.8%). For multimodal knowledge work — combining up to 1 million input tokens with up to 1 million output tokens to synthesize documents, spreadsheets, charts, audio and video — Google cites 53.4% on its Multimodal Knowledge Agent benchmark and 71.6% on Agents' Last Exam v1.1. In early partner testing on complex graphs, SQL and data-heavy enterprise workflows, Google says Hydrogen ran two to 2.8 times faster than 3.8 Flash while using 12% to 23% fewer tokens. All of these figures come from Google itself; no independent evaluation has been published yet.

What changes in Search

A separate Google AI blog post by Robby Stein, VP of Product for Google Search, describes the rollout as the largest model upgrade Search has received to date and confirms that Hydrogen is available globally to all users.

The post lists several concrete changes. Search now handles nuanced, multi-part and typo-ridden queries better, with deeper instruction following and tool calling so each part of a question is addressed. Google's query fan-out technique, which fires off simultaneous queries across the subtopics of a question, now surfaces a broader range of links. Responses are more visual, with more images, structured layouts, comparison tables and custom interactive tools, and answers improve in every language supported by AI Mode.

Why it matters

Hydrogen matters for two reasons. First, deployment scale: Google is putting a current-generation Gemini 4 model in front of every Search user worldwide, and its decision to use an efficiency-oriented model rather than a frontier one illustrates how heavily inference economics now shape products operating at Search's scale. Second, the release signals where Google believes AI products are heading — away from single do-everything models and toward routing work across a family of specialized models, coordinated by tools like Orchestrator. The flat $1.50 per million input tokens also puts direct pricing pressure on competitors' mid-tier models. The caveat is that every number in the announcement is Google's own, so independent benchmarks will show whether the claims hold.