Gemini 4 Hydrogen vs Carbon: What They Are and Which to Pick

Within three days, Google put two models from the Gemini 4 family on the table. On November 10, 2026, it made Gemini 4 Carbon, its most powerful model, generally available. On November 12, it introduced Gemini 4 Hydrogen, designed to be the "everyday" model for people working with AI agents. There are a lot of names, but the idea is simple: Carbon reasons deeply about hard problems, while Hydrogen runs fast and costs less.

Here you'll find what they are, how much they cost, where they're used, and when to pick one or the other. The test numbers are the ones reported by Google in its announcements: we pass them along, but they have not been verified by third parties.

Hydrogen and Carbon: what's the difference?

Gemini 4 HydrogenGemini 4 Carbon
What it's built forFast execution, agent workflowsHard, long tasks, deep reasoning
Announcement dateNovember 12, 2026November 10, 2026 (general availability)
Price per million tokens (input / output)$1.50 / $7.50$2 / $10 through November 30, then $4 / $20
Context windowUp to 1 million tokens in and 1 million outNot stated in the announcement

Google presents them as two tools that work together. In the Antigravity development environment, an "orchestrator" can switch from Carbon to Hydrogen within the same session: Carbon coordinates and decides, Hydrogen does the work and checks the results. According to the company, this gets task success close to Carbon's while using on average 24% less quota and finishing 19% faster.

What does Gemini 4 Hydrogen do?

  • Coding: according to Google, it makes targeted edits without rewriting unrelated files, and batches file reads. Partners report runs 2 to 2.8 times faster and 12-23% lower token use compared with Gemini 3.8 Flash.
  • Agent work: deep research, data analysis, and code exploration.
  • Documents and files: it accepts up to 1 million tokens in and produces up to 1 million out, to create reports from PDFs, spreadsheets, charts, audio, and video.
  • Reasoning levels: low, medium, high (default), and a new "xhigh" level.
  • App and search: it improves instruction following and tool use in the Gemini app, and makes Google Search answers more visual and rich in links.

What does Gemini 4 Carbon do?

Google calls it its strongest model, with frontier performance in cybersecurity, long agent workflows, coding, and scientific reasoning. It comes after Gemini 4 Argon, which remains the recommended choice for most coding and knowledge work. Carbon is for the toughest tasks, such as multi-step debugging or open-ended research, and it uses more tokens because it reasons and checks more.

How much do they cost and where are they used?

  • Hydrogen, developers: Gemini API and Google AI Studio (all tiers) and Android Studio; Google Antigravity for paid AI Pro and Ultra subscribers.
  • Hydrogen, businesses and consumers: Gemini for business, Gemini in Google Sheets (for eligible subscriptions), the model menu of the Gemini app, and Google Search. Free users of the Gemini app get it when automatic selection judges it suitable.
  • Carbon, developers and businesses: via API in Google AI Studio and Gemini for business, for all paying developers and businesses.
  • Carbon, users: coming soon for Google AI Ultra subscribers in Antigravity, in the Gemini app, and in Search's AI mode.
  • Carbon pricing: $2 per million input tokens and $10 output, the same as Argon, through November 30, 2026; after that date, $4 and $20.
  • Research and security: scientists can apply for six months of free access to Argon and Carbon at Antigravity's highest usage tier; Carbon for cybersecurity goes through the Fairwind program, with rolling applications.

The numbers reported by Google

These are tests chosen and measured by Google and its partners. They help you grasp the order of magnitude, not declare an absolute winner.

  • Hydrogen, Artificial Analysis index: 51 with default reasoning, versus 50 for Argon; it rises to 54 with "xhigh".
  • Hydrogen, Terminal-Bench 4.1: 57%, versus 18.2% for Gemini 3.8 Flash.
  • Hydrogen, compared with Gemini 3.8 Flash: tasks finished 2.5 times faster with default settings and more than 5 times fewer made-up claims.
  • Carbon: first on CursorBench 4.0, Perplexity's WANDR, and Zapier's AutomationBench, with a cost per task more than halved compared with the best external model.
  • Carbon, cybersecurity: in simulated attack tests it reports 99% of attempts blocked and a breach rate of 0.8% per attempt, versus 3.1% for Argon.

What changes for people at work?

For people who use AI to write, summarize documents, or prepare spreadsheets, the practical news is mostly Hydrogen: it costs less, answers faster, and if you use the free Gemini app you may already have it without noticing. Carbon makes sense if you need the best on complex, long problems, for example coding without supervision or analyzing many documents together, but remember that it uses more tokens and that the promotional price doubles at the end of the month. This suggestion is ours, drawn from the data in the announcements.

What isn't clear in the announcements?

  • The Carbon page doesn't state the context window and doesn't compare it directly with Hydrogen.
  • In the Hydrogen announcement, the text compares the model with Gemini 3.8 Flash, while an image caption cites Gemini 3.5 Flash: Google doesn't explain the difference.
  • All test results are chosen and reported by Google and its partners, with no independent verification.

Frequently asked questions

Are Hydrogen and Carbon the same model?

No, the announcements present them as different models with different roles: Carbon for deep reasoning, Hydrogen for fast execution. They can work together.

Can I use Gemini 4 Hydrogen for free?

According to Google, free users of the Gemini app get it through automatic model selection when it suits the request. For API use, the price is $1.50 per million input tokens.

Which one should I choose?

For everyday work, start with Hydrogen. Switch to Carbon, or to Argon, only if the task is truly hard and long.

Official sources

Transparency: this article is also written with the help of artificial intelligence models, starting from the official sources listed, and is reviewed before publication.

Condividi questo articolo sui tuoi socialShare this article on your social networksComparte este artículo en tus redes socialesTeile diesen Artikel in deinen sozialen Netzwerken

No products were found matching your selection.

Iscriviti alla newsletter di You vs AISubscribe to the You vs AI newsletterSuscríbete al boletín de You vs AIAbonniere den Newsletter von You vs AI