Google released Gemini 4 Hydrogen on Thursday, November 12, a faster and cheaper member of the Gemini 4 family that replaces Gemini 3.8 Flash across its products. It costs $1.50 per million input tokens and $7.50 per million output tokens across a 1 million-token context window. Google says Hydrogen scores 51 on the Artificial Analysis Intelligence Index at its default setting, one point above the larger Gemini 4 Argon. All benchmark figures in this story are Google’s own claims.
What Google released
Hydrogen is the execution model of the lineup. Google DeepMind describes Carbon and Argon as the models for long, difficult reasoning and Hydrogen as the one for fast, everyday coding and agent tasks. It accepts up to 1 million input tokens and, according to the announcement, can produce up to 1 million output tokens.
The model has four thinking levels: low, medium, high and a new xhigh tier. High is the default. Google says the index score rises to 54 at xhigh.
Unlike the two larger models, Hydrogen had no staged release. Ars Technica notes that Gemini 4 Carbon went first to security researchers, while Hydrogen was available immediately.
What Google claims against 3.8 Flash
Google positions Hydrogen as an upgrade for teams on Gemini 3.8 Flash. Its announcement lists these comparisons:
- 2.5 times faster task completion at the default high setting.
- 57 percent on Terminal-Bench 4.1, against 18.2 percent for 3.8 Flash.
- More than five times fewer fabricated claims.
- In early partner testing on data-heavy workloads, runs 2 to 2.8 times faster with 12 to 23 percent fewer tokens.
Google also cites results on newer evaluations: 58.8 percent on HLE-Diamond, 52.3 percent on the Vals Tax Agent Bench, 53.4 percent on its own new Multimodal Knowledge Agent benchmark and fifth place on Agent Arena. None of these has been independently reproduced yet, and one of the benchmarks was introduced by Google the same day.
On safety, Google reports a 96.7 percent honesty rate on an internal agent evaluation and says Hydrogen is more than 2.3 times more resistant to indirect prompt injection on the Gray Swan benchmark than its predecessor.
Price and availability
API pricing is flat across the whole context window. Ars Technica reports that the rate is unchanged from 3.8 Flash.
The model is rolling out globally in Google AI Studio, the Gemini API, Android Studio, Gemini for business and Gemini in Google Sheets. Paid Google AI Pro and Ultra subscribers can use it in the Antigravity coding environment.
In the Gemini app, Hydrogen appears in the model menu. A footnote in Google’s post says free users get it through automatic model selection when it suits the prompt. Ars Technica reports that free and AI Plus users otherwise remain on an older, lighter model.
Orchestrator and the Search upgrade
Two product changes arrive with the model. The first is Orchestrator in Antigravity, which switches between Carbon and Hydrogen inside one coding session. Google says that on Terminal-Bench 4.0 it reached task success close to Carbon alone while using 24 percent less quota and finishing 19 percent faster on average.
The second is Google Search. Hydrogen now powers AI responses in Search for all users worldwide, which Google calls its biggest model upgrade to Search so far. The company says Search handles multi-part questions better, shows more tables and images, and links to a wider range of sites through an improved query fan-out. Ars Technica points out that Google has promised broader linking before.
How Hydrogen compares with Argon and Carbon on cost
The price gap inside the family is about to widen. Carbon and Argon currently cost $2 per million input tokens and $10 per million output tokens, and they rise to $4 and $20 next month, Ars Technica reports. We covered the earlier tiers in our story on Argon’s rollout and pricing.
A worked example shows the difference. Take a coding agent that reads 20 million tokens and writes 2 million tokens in a day:
| Model | Input cost | Output cost | Daily total |
|---|---|---|---|
| Hydrogen | $30 | $15 | $45 |
| Argon or Carbon today | $40 | $20 | $60 |
| Argon or Carbon next month | $80 | $40 | $120 |
At next month’s prices the same workload costs about 2.7 times more on the larger models. That is the case for the split Google is promoting: plan with Carbon, execute and verify with Hydrogen. It only pays off if Hydrogen completes the routine steps reliably, which teams should check on their own tasks. Higher thinking levels also produce more output tokens, so the real bill depends on the setting.
What to watch next
- Independent results from Artificial Analysis and Terminal-Bench maintainers that confirm or adjust Google’s figures.
- The Carbon and Argon price increase scheduled for next month.
- Whether free Gemini app users get direct access to Hydrogen in the model menu.
- Publisher traffic data after the Search change to query fan-out and linking.
Key takeaways
- Gemini 4 Hydrogen replaces Gemini 3.8 Flash across Google’s developer, enterprise and consumer products.
- It costs $1.50 and $7.50 per million input and output tokens, flat across a 1 million-token context.
- Google claims a score of 51 on the Artificial Analysis Intelligence Index at the default setting and 54 at xhigh.
- A new Orchestrator in Antigravity switches between Carbon and Hydrogen within a session.
- Hydrogen now powers AI responses in Google Search for all users.
Frequently asked questions
How much does Gemini 4 Hydrogen cost?
The API price is $1.50 per million input tokens and $7.50 per million output tokens. The rate is the same across the full 1 million-token context window.
Is Gemini 4 Hydrogen free in the Gemini app?
Google says free users get Hydrogen through automatic model selection when it fits the prompt. Paid subscribers can choose it directly in the model menu.
What is the difference between Gemini 4 Hydrogen and Argon?
Hydrogen is the faster, cheaper model for everyday coding and agent tasks. Argon and Carbon are Google’s larger models for long, difficult reasoning and cost more per token.
