# Gemini 4 Hydrogen arrives in Google Search worldwide, with benchmark gains that remain Google's own claims

Published 2026-11-13T09:01:03.000Z · AI Pulse · https://ai-pulse.news/models-labs/gemini-4-hydrogen-google-search-rollout

*This story was written with the help of artificial intelligence, checked automatically, then reviewed by an editor before publication, under our AI policy : https://ai-pulse.news/ai-policy*

On 12 November 2026, Google DeepMind announced Gemini 4 Hydrogen, a speed-focused model now deployed in Google Search for all users worldwide and priced at $1.50 per million input tokens and $7.50 per million output tokens.

## Key points

- Google DeepMind announced Gemini 4 Hydrogen on 12 November 2026, alongside the frontier models Gemini 4 Carbon and Gemini 4 Argon.
- Hydrogen is rolling out in Google Search for all users worldwide and is the new default model in AI Mode, according to Google Search product VP Robby Stein.
- API pricing is $1.50 per million input tokens and $7.50 per million output tokens, with the same price across the one-million-token context window.
- Performance figures, including an Intelligence Index score of 51 against Argon's 50, come from Google and have not been independently confirmed in the material available.

## What Google announced

Google DeepMind introduced Gemini 4 Hydrogen on 12 November 2026. According to the [DeepMind announcement](https://deepmind.google/blog/introducing-gemini-4-hydrogen/), the model is oriented towards speed and efficiency for everyday coding and agentic workflows. It comes alongside two frontier models, Gemini 4 Carbon and Gemini 4 Argon.

DeepMind frames the release as broad access: "Today, we’re introducing Gemini 4 Hydrogen, bringing cutting-edge intelligence to everyone."

Google lists availability from the day of the announcement in the Gemini API, Google AI Studio, Android Studio, Gemini for business, the Gemini app, Google Search and Google Antigravity. In Antigravity, access is for paying Google AI Pro and Ultra subscribers.

## Hydrogen in Google Search

Hydrogen is being deployed in Google Search for all users worldwide, according to [Google's Search blog](https://blog.google/products-and-platforms/products/search/gemini-4-hydrogen-search-upgrade/). The company says it improves understanding of complex, multi-part questions, produces more visual answers (images and comparison tables) and uses an expanded query fan-out technique. Google's own description: "This is the biggest model upgrade to Search yet." That is a company characterisation, not an independent assessment.

Robby Stein, VP of Product for Google Search, told [The Verge](https://www.theverge.com/ai-artificial-intelligence/1032505/google-search-ai-mode-gemini-4-hydrogen) that Hydrogen is "now the new default model in AI Mode". He said it is also used for the hardest questions in AI Overviews, which implies it does not handle every AI Overviews query.

## Pricing and benchmark claims

Google sets Hydrogen's price at $1.50 per million input tokens and $7.50 per million output tokens. The same price applies across the full one-million-token context window.

The performance figures below are reported by Google unless stated otherwise.

- **Intelligence Index:** Google says Hydrogen surpasses Argon on the Artificial Analysis Intelligence Index, scoring 51 against 50 at the default setting and 54 at the xhigh setting.
- **Speed and coding:** Google says Hydrogen finishes tasks 2.5 times faster than Gemini 3.8 Flash at its default high setting, and lifts its Terminal-Bench 4.1 score from 18.2% to 57%.
- **Reliability:** Google says Hydrogen cuts fabricated claims by more than 5 times compared with 3.8 Flash. It also reports an 81% reduction in agentic task overreach and a 79% reduction in underservice, a 12-fold gain in reducing misaligned actions, and a 96.7% honesty rate on its internal evaluation.
- **Other results:** Hydrogen ranks 5th on Agent Arena, the live benchmark from arena.ai, and scores 58.8% on HLE-Diamond, according to Google. ZDNet France also reports the fifth-place Agent Arena ranking, and a 52.3% score on Vals Tax Agent Bench that Google describes as first place.

A one-point gap on the Intelligence Index (51 versus 50) is small, and the announcement material does not describe the margin of error. The honesty rate comes from an internal evaluation, so it cannot be compared directly with other models' results.

## Agentic use and competitors

Google says Orchestrator, in Google Antigravity, switches dynamically between Carbon and Hydrogen within a single session. On Terminal-Bench 4.0, it reportedly reaches success rates close to Carbon while using on average 24% less quota and running 19% faster. This is a company claim about its own tooling.

According to The Verge, Google compares Hydrogen with Claude Sonnet 5.5 and GPT-6.1 Terra, and says it matches or beats some competing models in its class. The material available does not give the competitor scores or the testing conditions, so the extent of that lead is not established here.

### Open questions

The announcements do not say how Hydrogen performs on independent evaluations beyond the ranking on Agent Arena. They also do not detail how Search chooses between Hydrogen and other models for AI Overviews queries.

**Gemini 4 Hydrogen: key figures reported by Google** (data as of 2026-11-12)

| Metric | Reported value |
| --- | --- |
| Input price | $1.50 per million tokens |
| Output price | $7.50 per million tokens |
| Context window | 1 million tokens |
| Intelligence Index (default / xhigh) | 51 / 54 (Argon: 50) |
| Terminal-Bench 4.1 (vs 3.8 Flash) | 57% (from 18.2%) |
| HLE-Diamond | 58.8% |
| Internal honesty rate | 96.7% |

## FAQ

### What is Gemini 4 Hydrogen?

A Google DeepMind model oriented towards speed and efficiency for everyday coding and agentic workflows, announced on 12 November 2026 alongside Gemini 4 Carbon and Gemini 4 Argon.

### How much does Gemini 4 Hydrogen cost?

Google lists $1.50 per million input tokens and $7.50 per million output tokens, at the same price across the one-million-token context window.

### Is Hydrogen the default model in Google Search?

It is the new default model in AI Mode, per Robby Stein. In AI Overviews it is used for the most difficult questions.

### Where can developers and users access it?

Google lists the Gemini API, Google AI Studio, Android Studio, Gemini for business, the Gemini app, Google Search and Google Antigravity (for paying Google AI Pro and Ultra subscribers).

## Sources

1. [Introducing Gemini 4 Hydrogen](https://deepmind.google/blog/introducing-gemini-4-hydrogen/), Google DeepMind, 2026-11-12 (primary source)
2. [Google Search gets a major upgrade with Gemini 4 Hydrogen](https://blog.google/products-and-platforms/products/search/gemini-4-hydrogen-search-upgrade/), Google, 2026-11-12 (primary source)
3. [Google’s new Gemini model is better at understanding complicated questions](https://www.theverge.com/ai-artificial-intelligence/1032505/google-search-ai-mode-gemini-4-hydrogen), The Verge, 2026-11-12
4. [Google lance Gemini 4 Hydrogen, son modèle rapide pour le code et les agents, dans Search et l’application Gemini](https://www.zdnet.fr/actualites/google-lance-gemini-4-hydrogen-son-modele-rapide-pour-le-code-et-les-agents-dans-search-et-lapplication-gemini-507821.htm#xtor=RSS-1), ZDNET France, 2026-11-13
