Google Makes Gemini 4 Hydrogen the Default Model for AI Mode Search
Google's Gemini 4 Hydrogen is now the default AI Mode model, breaking down multi-part queries, adding 'xhigh' thinking and richer visuals across Search, Gemini and AI Studio.
- Gemini 4 Hydrogen became the default model in AI Mode and powers hardest questions in AI Overviews, Google announced today.
- The model is available now in Search, the Gemini app, all AI Studio tiers, the Gemini API, Gemini for Business, and Antigravity for AI Pro/Ultra subscribers.
- It adds a new 'xhigh' thinking level alongside low, medium, and high.
- Google claims it beats Gemini Flash 3.8 in speed and intelligence and matches or beats Claude Sonnet 5.5 and GPT-6.1 Terra — vendor claims, not independent benchmarks.
- Responses now include simultaneous subtopic queries with wider link sets plus images, tables, graphs and maps.
Google confirmed today that Gemini 4 Hydrogen is now the default model powering AI Mode in Search and handles the hardest questions surfaced in AI Overviews. Robby Stein, VP of product at Google Search, said the model is "the biggest model upgrade to Search yet."
The rollout is broad. As of today, Gemini 4 Hydrogen runs in Search, the Gemini app, all tiers of Google AI Studio, the Gemini API, Gemini for business, and in Google Antigravity for Google AI Pro and Ultra subscribers. That means publishers and SEOs tracking AI Mode responses and AI Overviews are now looking at a materially different answer engine than the one deployed yesterday.
What changes in how Search answers questions?
Google says the new model gives more helpful answers to complex questions — specifically "questions with multiple parts, hard-to-understand typos, or highly specific requests." The model breaks a query into its component parts and addresses each one in its response.
For search practitioners, that's a meaningful shift in query handling. Multi-part and long-tail queries have historically fragmented into several searches; Hydrogen is built to resolve them in a single AI Mode session.
The model also widens link selection. Google says it now conducts "simultaneous queries across all the relevant subtopics," which returns a broader set of sources inside each response. Stein's framing: more subtopic coverage, more citations per answer.
What do the new responses look like?
Responses carry more visuals, Google confirmed — images, comparison tables, graphs, and maps. One example from Stein's demonstration shows the AI generating an interactive animated driving simulation to answer a user's question about how phones locate which highway lane a car occupies in areas without cell service.
That example signals where Google is heading: answers as interactive artifacts, not just text with links. Comparison tables and graphs inside AI Mode directly compete with publisher comparison content, a category already under pressure from AI Overviews.
Language coverage also improves. Google says the model delivers better language understanding "across all supported AI Mode languages" — relevant for international SEO teams watching how AI Mode expands beyond English markets.
What thinking levels are available?
Users now get four thinking-level options: low, medium, high, and a newly added "xhigh" mode. Higher thinking levels typically mean deeper reasoning and longer processing before an answer renders — Google did not specify which level AI Mode uses by default for hardest-question routing in AI Overviews.
How does it compare to rival models?
Google's stated benchmarks position Hydrogen against top-tier competitors. The company says the model "delivers significant leaps in both speed and intelligence" compared to Gemini Flash 3.8, and that it "matches or outperforms some of the competing models in its class," naming Claude Sonnet 5.5 and GPT-6.1 Terra specifically.
Note the sourcing line here: these are Google's own claims, not independent benchmark results. No third-party evaluation accompanied the launch.
Which claims are confirmed, and which are Google's say-so?
Confirmed by Google's announcement:
- Gemini 4 Hydrogen is live today across Search, the Gemini app, AI Studio (all tiers), the Gemini API, Gemini for Business, and Antigravity for AI Pro and Ultra subscribers.
- It is the default model in AI Mode and powers the hardest questions in AI Overviews, per Stein.
- It adds "xhigh" as a fourth thinking level.
Unverified, Google-asserted only:
- The "significant leaps in both speed and intelligence" over Gemini Flash 3.8.
- Parity or superiority versus Claude Sonnet 5.5 and GPT-6.1 Terra.
- Improved language understanding across all supported AI Mode languages.
What should publishers and SEOs watch now?
Watch AI Mode and AI Overview responses over the coming days for three things: which subtopics now get cited under the simultaneous-query approach, whether comparison tables and graphs displace publisher equivalents in commercial verticals, and how the harder-question routing changes which sites earn links from AI Overviews. The change is live now, so the diff starts today.
via theverge.com (Source)
More from James Calloway
Show full bio
Market editor covering money, funding and market shifts around Search rankings and algorithm updates, generative engine optimization at SERP Journal.
220 articles