Google introduced Gemini 4 Hydrogen on November 12, 2026, extending its latest model family to the everyday tasks that many people use most: questions, writing, search, coding, and multi-step work. Unlike the more restricted Carbon rollout earlier in the week, Hydrogen is intended to reach free and paid users across Google products. There is one important catch: having access does not always mean you can pick Hydrogen manually in every interface. Google's official announcement
The short answer is that Google says Hydrogen is available through Google AI Studio, the Gemini API, the Gemini app, Search and business products as rollout reaches eligible users. In the free Gemini app, its official footnote specifically describes automatic model selection when appropriate; paid-user interfaces and Google Antigravity follow different conditions. Developers also get a published model price of $1.50 per million input tokens and $7.50 per million output tokens across a one-million-token context window. Those are usage prices, not monthly Gemini subscription fees.
Where can you use it?
| Surface | What Google announced | What you should not assume |
|---|---|---|
| Gemini app | Global rollout; free users can receive Hydrogen through automatic selection | Every free account has a manual Hydrogen picker |
| Google Search | A Hydrogen-backed upgrade rolling out globally across supported AI experiences | Every search result always invokes Hydrogen or every feature works in every country |
| Google AI Studio | Availability for all Studio tiers | Every API use is free or unconstrained |
| Gemini API | Hydrogen offered to developers | Pricing is the same as a consumer subscription or quotas never apply |
| Google Antigravity | Hydrogen for eligible paid Google AI Pro/Ultra subscribers; Orchestrator introduced | Orchestrator is a separately proven productivity guarantee |
| Business integrations | Gemini for business, eligible Sheets integrations and API deployment | A company's account, plan, compliance region and admin settings need no checks |
This matrix is a summary of Google's November 12 announcement and its explicit app footnote, not a hands-on regional entitlement test. If the option is missing, check your current product surface, account tier, supported location and rollout status rather than assuming the official announcement promised an identical switch to everyone. Official model announcement
What does Hydrogen cost through the API?
Google lists a flat token price across the model's one-million-token context window:
| Published API unit | USD per 1,000,000 tokens |
|---|---|
| Input tokens | $1.50 |
| Output tokens | $7.50 |
Input represents the material sent to the model; output represents generated tokens that may include reasoning or other billed output under the applicable product rules. The context window describes how much information a request can accommodate; it is not a free monthly allowance, guaranteed maximum output, or a price for one entire conversation. For example, at those published token rates, a hypothetical request with 200,000 billed input tokens and 10,000 billed output tokens would correspond to $0.30 + $0.075 = $0.375 before any other billable features or terms. That is an illustration, not a tested invoice or a quote for a particular account. Confirm today's API pricing, rate limits and caching/batch terms before production adoption. Google price statement
For casual users, those API rates should not be confused with what a Gemini app subscription costs. A free app experience also does not mean that a programmatic API workload has unlimited free capacity.
Hydrogen, Carbon and Argon: which one is which?
Google positions Hydrogen as a fast execution model for coding, tool use and everyday agent tasks. Carbon and Argon are the models it emphasizes for harder, longer-horizon reasoning. That is a vendor positioning statement, not proof that one model is universally better.
There are two useful decisions. First, a regular Gemini app or Search user can simply try an eligible prompt and check the visible behavior rather than worrying about a model benchmark. Second, a developer who runs frequent code edits or retrieval-driven steps can compare Hydrogen's cost and latency against a stronger reasoning model with their own tasks. Hard debugging and architectural questions may still justify another model even when the per-token price is higher.
Google reports an Artificial Analysis Intelligence Index score of 51 for Hydrogen at the default high thinking level, compared with 50 for Argon, and 54 at Hydrogen's highest xhigh setting. Scores from different thinking levels or test setups cannot be treated as a single direct price-adjusted ranking. Google also says Hydrogen can be 2.5 times faster than Gemini 3.8 Flash at its high default setting on the tasks it evaluated. HotRank has not benchmarked these claims. Google performance discussion.
What is Antigravity Orchestrator?
Orchestrator is not just another model name. Google describes it as a system that switches dynamically between models such as Carbon and Hydrogen during a workflow. The intended pattern is to put long-form planning in a stronger model and routine code editing, testing or verification in a faster one. The same announcement also describes parallel workers in isolated workspaces; not all products share identical orchestration features.
Google reports 24% less quota and 19% faster completion while approaching Carbon's success rate. Its sources disagree on the benchmark version: the model announcement labels this Terminal-Bench 4.0, while the Antigravity team's detailed post labels it 4.1. We cannot resolve that discrepancy; the figures should not be presented as one independently verified comparison. The figures are Google's own test claims, not a guarantee of savings for your repository. Results can change with prompt composition, tool permissions, parallelism, retries and quotas. Official Orchestrator description
It is also distinct from the /swarm preview announced alongside the earlier Carbon rollout: both involve coordination ideas, but they are not interchangeable product names or proof that all orchestration options have identical access.
What changes in Google Search?
Google separately says the Search experience is being upgraded with Hydrogen to handle multi-part queries, find relevant sources and produce more visual responses such as tables and structured answers. The Search team calls the rollout global for users of its supported experiences and languages. That is Google's announcement of a product upgrade, not an independent measurement that every answer has become more accurate or that every search is generated by the same model. Google Search announcement
For readers, a useful test is to ask a specific, multi-constraint question, inspect the citations and compare the information against original sources. A polished table still needs fact checking; visual presentation alone is not proof of correctness.
How should you try Hydrogen now?
If you use the Gemini app: open the official Gemini app and inspect any model choices available to your account. On a free account, Google's footnote says Hydrogen may be selected automatically when it fits your prompt; absence from a manual dropdown does not alone prove absence of access. Do not purchase a plan solely to make the model appear without checking the current plan page.
If you build with Google AI Studio or the API: open Google AI Studio and start with a short reproducible workload. Record model identifier, thinking level, input/output usage, response quality, latency and cost under the same prompt. Test simple tasks and your hardest task separately; count retries and tool failures, not only successful output. No test in this article has been performed by HotRank.
If you use Antigravity: verify that your account is an eligible paid Google AI Pro or Ultra subscriber and that Orchestrator is offered in your version. Its advertised efficiency figures are a reason to test a workflow, not a promise of an automatic bill reduction.
What remains unverified?
As of our November 15 editorial check, we cannot verify each country's live feature toggle, every free/paid account's exact picker state, rate-limit policies, separate caching or batch discounts, a universal manual-selection entitlement, or independent performance and safety benchmarks. Google's article was published November 12 and marked updated November 13. Prices and entitlements can change; check the links below for the latest applicable rules.
Sources and update triggers: Google's launch and pricing announcement; Google's Search rollout note. Update this story when Google changes access tiers, API prices, Orchestrator permissions or documented rollout. This is a source-reviewed news explainer, not a product hands-on test.
For Orchestrator, the official starting point is Google Antigravity; its team describes switching back to Carbon when Hydrogen encounters friction, rather than routing the whole task once. For more reporting, visit our AI coverage or latest stories.
Visual: the Gemini brand logo, not a Hydrogen interface screenshot. Gemini and Google are trademarks of Google LLC. HotRank is not affiliated with or endorsed by Google.
