Gemini 4 Hydrogen: Check the Cost Before Switching

Google introduced Gemini 4 Hydrogen on November 12, 2026, as an execution model for coding and agent workflows. Its Standard API rate is $1.50 per million input tokens and $7.50 per million output tokens, including thinking tokens. At the same uncached token count, that is twice the current Standard price of Gemini 3.8 Flash. A faster task does not automatically mean a smaller API bill.

What changed on November 12

Google positions Hydrogen as a path for teams using 3.8 Flash for everyday code writing, testing and agent execution. The company says it finishes tasks 2.5 times faster than Flash at Hydrogen’s default high thinking setting and reports higher scores on its cited coding benchmark. Those are Google’s reported results, not a test of your repository, prompts or tools.

Hydrogen is available through Google AI Studio and the Gemini API, among other Google products, according to the announcement. The API model code is gemini-4-hydrogen. Google lists low, medium, high and xhigh thinking levels. A migration should specify the level used for the comparison: changing both the model and the thinking setting makes the result harder to interpret.

The DailyTech model tracker now lists the release. Use Google’s documentation for the limits and rates that go into a buying decision.

Compare the bill at the same token count

These are Google’s published Standard paid-tier rates in US dollars per million tokens. Output prices include thinking tokens. The pricing page lists a scheduled Flash rate change on January 1, 2027.

Model and period	Input / 1M tokens	Output / 1M tokens
Gemini 3.8 Flash, through Dec. 31, 2026	$0.75	$3.75
Gemini 4 Hydrogen, current Standard rate	$1.50	$7.50
Gemini 3.8 Flash, from Jan. 1, 2027	$1.50	$7.50

At today’s rates, Hydrogen doubles both uncached token rates. From January 1, the published Flash rates match Hydrogen’s current rates. This comparison holds the billing tier and token mix constant; it is not a prediction of your actual bill. Google also offers Batch and Flex rates, but mixing those with a Standard price would distort the comparison.

The pricing page separately lists cached-input rates. Hydrogen’s Standard implicit-cache rate is $0.075 per million cached tokens. Flash currently lists $0.075 per million cached tokens, plus a separate storage charge for explicit caching; its published cached-input rate rises to $0.15 on January 1. Cache hit rates and storage usage therefore need their own calculation. Our prompt caching explainer covers the mechanics behind that part of the bill.

A million-token context is a shared budget

Google’s Hydrogen model page lists a unified 1,048,576-token input-and-output context. Its maximum output is 65,536 tokens for a default synchronous request. Google’s optional long-decode route can extend output up to 1,048,576 tokens. The million-token figure is a combined budget, not a promise of one million input tokens followed by another million output tokens. Google’s launch announcement describes million-token input alongside output of up to a million tokens. Its API specification explicitly defines a shared input-and-output limit. Use the API specification when sizing a request.

That distinction matters for long repository scans, transcripts and large agent histories. Leave room for the model’s response and thinking tokens when you size the prompt. The default output cap still applies unless you use the documented long-decode path. For a broader explanation of why an answer needs space inside the window, see our context-window guide.

Hydrogen supports implicit caching only; Google’s model and pricing pages say explicit caching is unsupported. If your Flash integration creates named caches or depends on explicit cache storage, do not assume that code and billing behavior transfer unchanged. Inventory those calls before changing the model identifier.

Price a migration before changing production

Here is an illustrative arithmetic example, not a measured workload. Suppose a month of requests uses 10 million uncached input tokens and 2 million billable output tokens, with the same token counts on both models. The formula is:

input millions × input rate + output millions × output rate

Standard paid-tier scenario	Calculation	Illustrative token charge
3.8 Flash through Dec. 31	10 × $0.75 + 2 × $3.75	$15
4 Hydrogen today	10 × $1.50 + 2 × $7.50	$30
3.8 Flash from Jan. 1	10 × $1.50 + 2 × $7.50	$30

The example excludes cached tokens, explicit-cache storage, grounding and other tool charges, retries, taxes and infrastructure. It assumes the same numbers of successful tasks and billable tokens. Use the API cost calculator with your own inputs, then check the result against the current Google price table.

At today’s uncached rates, Hydrogen would need to use less than half as many billed tokens to cost less on tokens alone if the input-to-output mix stays the same. If the mix changes, calculate the two weighted bills directly. A model that completes more tasks successfully may still be worth a higher token bill; calculate cost per successful task rather than treating speed or unit price as the whole decision. Our Carbon pricing analysis shows that same comparison method for a different Google model.

Decide with a small representative evaluation

Choose a sample of real tasks that includes easy requests, long inputs and cases where tools fail. Run Flash and Hydrogen against the same tasks, record the thinking level and count successful completions. Measure elapsed time, input tokens, output and thinking tokens, retries, tool errors and cached-input tokens. Then multiply each billing category by the relevant published rate. This is a suggested evaluation method; we did not run these models or measure their performance.

Check code paths that use explicit caching before switching. Also check whether any request relies on outputs above the default 65,536-token synchronous cap or on a particular context allocation. If the results meet your success and latency targets at an acceptable cost per successful task, migrate gradually and monitor the same measures in production. Otherwise keep Flash while you adjust prompts, caching or the workload split.

Sources

• Google’s November 12 announcement — release date, positioning, availability and Google-reported performance.
• Gemini API pricing — Hydrogen and Flash Standard token and cache rates, including the scheduled Flash change.
• Gemini 4 Hydrogen model specification — model code, context and output limits, caching support.

All pages read on November 13, 2026.

Free toolWhat will 10,000 chats a month cost?Estimate my bill