Google has launched Gemini 4 Hydrogen, a new model built for speed, and put it at the heart of its search engine. The company says Hydrogen is now the default model in AI Mode and answers the hardest questions in AI Overviews, reaching every Search user worldwide from today. Google calls it the biggest model upgrade to Search yet, and the launch comes just two days after its flagship sibling, Gemini 4 Carbon, arrived for paying subscribers. Hydrogen is also available through the Gemini API, the Gemini app, and Google’s developer and business tools.
What Hydrogen is
Hydrogen is the fast member of the Gemini 4 family. Google positions its frontier models, Carbon and Argon, for deep, long-horizon reasoning on the hardest problems, while Hydrogen is built for rapid execution across everyday coding and agentic workflows. It is the model that does the work quickly, rather than the one that thinks the longest.
The model comes with four thinking levels — low, medium, high, and a new xhigh tier — so users can trade reasoning depth against speed and cost. At its default high setting, Hydrogen already outscores Argon on the Artificial Analysis Intelligence Index, 51 to 50, and reaches 54 at its xhigh setting, while finishing tasks faster than models in its class.
The speed claims come with numbers. Hydrogen finishes tasks two and a half times faster than Gemini 3.8 Flash, while more than tripling its Terminal-Bench score, from 18.2 percent to 57 percent, and cutting fabricated claims by more than five times. It also ranks fifth on Agent Arena’s live benchmark and leads its class on several reasoning and enterprise tests.
Pricing is flat across the model’s one-million-token context window: $1.50 per million input tokens and $7.50 per million output tokens. Developers can use Hydrogen through the Gemini API and AI Studio, a day after Google launched a separate Decisions API for fast, closed-set calls. The flat rate contrasts with Argon, whose price doubles on December 1.
For Search users, the change shows up four ways. Google says Hydrogen understands intent better, even for multi-part questions, typos, and specific requests. Its improved query fan-out runs simultaneous searches across subtopics, linking to more sites. Responses are more visual, with more images, comparison tables, and custom tools, and richer language understanding extends across every supported AI Mode language.
The upgrade landed in the middle of Google’s November core update, which began rolling out on 10 November and can run for up to two weeks. That timing matters: the model change reaches users while the ranking systems underneath Search are still shifting, so early reactions to either change will be hard to separate.
Google is also using Hydrogen to power a new multi-model trick. Orchestrator, inside Google Antigravity, switches between Carbon and Hydrogen within a single session, letting Carbon plan while Hydrogen executes. In Google’s tests, the combination reached near-Carbon success rates on Terminal-Bench while using 24 percent less quota and finishing 19 percent faster.
Google also leans on safety numbers for a model built to act on its own. Hydrogen cuts overreach in agentic tasks by up to 81 percent and misaligned actions by twelve times, and it is more than twice as resilient to indirect prompt-injection attacks. The company says the model preserves 99.9 percent chain-of-thought legibility across more than 200,000 audited turns.
Our Read
Read plainly, Hydrogen is Google’s answer to a problem every model maker now faces: frontier models are powerful but slow and expensive, and most daily work does not need their full depth. Hydrogen is the workhorse — fast enough for every Search user, cheap enough to price flat, and strong enough to beat its bigger siblings on several tests.
The Search angle is the one that matters most. A model upgrade reaching every Search user on day one is distribution no rival can match overnight, and Google is using it to push AI Mode and AI Overviews into everyday queries. The question is whether users notice — better fan-out and more answers help if they lead to clicks, not summaries.
The honest caveat is that every number here is Google’s own. The benchmark leads, the speed multiples, and the safety gains all come from Google’s blog post and model card, not from independent testing. Hydrogen’s test will be whether developers and Search users feel the speed daily, and whether its answers hold up against the models it claims to beat.
Why It Matters
Hydrogen matters because it changes what the default AI in Search can do. When the model answering billions of queries gets faster and better at understanding intent, every search becomes a little more capable — and every site that depends on search traffic feels the shift in what gets linked and surfaced.
It also sets the pattern for how Google will ship Gemini 4: frontier models for the hardest work, fast models for everything else, with the two working together inside tools like Antigravity. Rivals now have to match not just one model, but a whole system of models that switch roles mid-task.
Background
Hydrogen arrives two days after Gemini 4 Carbon reached paying subscribers in AI Mode, and in the middle of Google’s November core update, which began rolling out on 10 November. The Gemini 4 family now spans Carbon and Argon for frontier reasoning and Hydrogen for fast execution, with a new Intelligence Index lead to show for it.
Source: Google’s Hydrogen announcement and Search Engine Roundtable’s report.
0 Comments