What happened

Google announced Gemini 4 Hydrogen on Thursday, November 12, and rolled it out the same day — the new version of its fast, everyday AI model. It replaces Gemini 3.8 Flash across Google's products, including Search, the Gemini app and the developer API.

Hydrogen continues Google's elemental naming scheme, following Argon and Carbon. But unlike those two, which rolled out slowly to security researchers and top-tier subscribers first, Hydrogen is available right away. Everyone gets it in Search. In the Gemini app and AI Studio, it is live now for AI Pro and Ultra subscribers. Free and AI Plus users stay on the older 3.5 Flash-Lite model for now.

Google says Hydrogen is a major step up from Flash 3.8. It runs at four thinking levels: low, medium, high, and a new xhigh setting. Even in high mode, it is 2.5 times faster than the model it replaces. Google also claims it cuts fabricated claims by more than five times, based on its own analysis.

The announcement at a glance

  • Announced: Thursday, November 12, 2026, with the rollout starting the same day — no slow, staged release this time.
  • What it replaces: Gemini 3.8 Flash, across Search, the Gemini app and the developer API.
  • Who gets it now: everyone in Google Search; AI Pro and Ultra subscribers in the Gemini app and AI Studio. Free and AI Plus users stay on 3.5 Flash-Lite for now.
  • Speed: 2.5 times faster than 3.8 Flash in high mode, with four thinking levels — low, medium, high and a new xhigh.
  • Accuracy claim: Google says fabricated claims drop by more than five times, based on its own analysis.
  • Price: unchanged for developers — $1.50 per million input tokens and $7.50 per million output tokens.

Why it matters

Speed and accuracy are the two things that matter most in a model people use all day. Flash models power the quick answers in Search and the Gemini app, so a faster, more reliable version changes what millions of people see every time they ask a question.

The coding gains are the sharpest part. Google says Hydrogen more than triples 3.8 Flash's score on Terminal-Bench, a test of real coding tasks. On the Artificial Analysis Intelligence Index, Hydrogen scored 54 in xhigh mode, ahead of Argon at 50. That puts a fast, cheap model within reach of Google's own top-tier offering.

It also shows how fast the AI race is moving. [Anthropic launched its Mythos and Fable 5.2 models](/article/anthropic-launches-mythos-fable-5-2-models) days ago, and [Apple is racing to turn Siri into a full AI agent](/article/apple-races-to-make-siri-ai-agent). Google is answering with a model that is both quicker and cheaper to run. The same stretch brought [ChatGPT's wallet, which lets AI agents shop for you](/article/chatgpt-wallet-lets-ai-agents-shop-for-you), and [LangChain's $3 billion valuation in the AI agent race](/article/langchain-hits-3b-valuation-ai-agent-race) — every major AI player is shipping at once.

Hydrogen vs. Argon and Carbon

Hydrogen is the fast, everyday member of the Gemini 4 family. Argon and Carbon are positioned as Google's best models for long, complex tasks and security-sensitive work, while Hydrogen is aimed at simpler code, quick tests and the agentic coding tools developers use daily, where speed matters more than depth.

On benchmarks, Hydrogen holds its own against the bigger models: it scored 54 on the Artificial Analysis Intelligence Index in xhigh mode, ahead of Argon's 50. It is also the cheaper option for developers. API pricing is unchanged at $1.50 per million input tokens and $7.50 per million output tokens, while Carbon and Argon cost $2 and $10 today and rise to $4 and $20 next month. Unlike Argon and Carbon, which rolled out slowly to security researchers and top-tier subscribers first, Hydrogen was available immediately.

What it means for the US

For US readers, the change shows up first in Google Search. Google says this is the biggest update ever to its AI-powered search tools. Hydrogen is built to handle long, multi-part questions better, and answers in AI Overviews and AI Mode will carry more visual elements. Google also says its fan-out search, which pulls from many sources at once, is now tuned to link to more of them.

Developers get the same pricing as before: $1.50 per million input tokens and $7.50 per million output tokens through the API. That matters because the two bigger models, Carbon and Argon, cost $2 and $10 today and jump to $4 and $20 next month. Hydrogen gives US startups and developers a cheaper way to build AI features without paying flagship prices.

What happens next

Google positions Argon and Carbon as its best models for long, complex tasks and security-sensitive work. Hydrogen is aimed at simpler code, quick tests and the agentic coding tools developers use daily, where speed matters more than depth and token budgets are tight.

Watch whether the free tier gets Hydrogen next. Google has not said when AI Plus and free users move off 3.5 Flash-Lite. The bigger test is the accuracy claim. If fabricated claims really drop as much as Google says, it will show up fast in Search answers.