KovAPIKovAPI
Chat ● Operational Service status →

Gemini 4 Hydrogen

Beta ⧉
gemini-4-hydrogen

Gemini 4 Hydrogen is the Gemini 4 model Google released on November 12, 2026, listed at $1.50 input and $7.50 output per million tokens, the lowest price in the Gemini 4 family. It is available in Beta for text chat and reasoning with three thinking levels, Low, Medium and High; Low is the default and thinking tokens bill at the output rate. The official 1,048,576-token context is shared by input and output. Image input, tool calling and JSON structured output are supported. Repeated long inputs produced implicit cache hits, billed at one twentieth of the input rate; a hit on every repeat is not guaranteed.

BetaLow / Medium / High1M contextOpenAI compatible

Live Test · Playground

Try out Gemini 4 Hydrogen right here (available after login).

Input

Advanced
Web search
Memory · multi-turn

Conversation

Start a conversationType a message below to begin

About Gemini 4 Hydrogen

Gemini 4 Hydrogen is the Gemini 4 model Google released on November 12, 2026, available on KovAPI in Beta for text chat and reasoning through the OpenAI-compatible API. Its list price is $1.50 input and $7.50 output per million tokens, the lowest in the Gemini 4 family. Thinking comes in three levels, Low, Medium and High; Low is the default and thinking tokens bill at the output rate. The official 1,048,576-token context is shared by input and output. Image input, tool calling and JSON structured output are supported, and repeated long inputs produced implicit cache hits billed at one twentieth of the input rate.

Use cases

Reasoning at a lower price

Math, logic and coding problems that need thinking at volume: Hydrogen's list price is a quarter below Argon and Carbon, and you can set thinking from Low to High to match each problem.

Questions over long documents

Put a whole manual, a long report or a chat history into the million-token context and ask directly, without building a retrieval step first.

Image understanding

Read charts, count objects and pick out letters and colours in a grid, with answers that show their reasoning.

Tool calling and JSON output

Function calling and json_schema structured output both work, which suits batch extraction, classification and decision steps in automated pipelines.

How to choose

Pick Hydrogen when you want Gemini 4 at the lowest cost. Pick Gemini 4 Argon when long, multi-step reasoning matters more. For faster, cheaper simple jobs at volume, use Gemini 3.8 Flash.

FAQ

How do I set the thinking level?
Send reasoning_effort="low", "medium" or "high" in the OpenAI-compatible request, or pick it in Advanced settings on this page. Default: low. All levels use the same token rates; more thinking consumes more output tokens.
How is it different from Gemini 4 Argon?
Hydrogen has the lowest list price in the Gemini 4 family ($1.50 / $7.50 versus Argon at $2 / $10). Choose Argon for long, multi-step reasoning and Hydrogen when cost matters more.
How large is the context?
The official window is 1,048,576 tokens shared by input and output. Split very long jobs during Beta.
Does it support images and tool calling?
Yes. Send images in the message, or define tools in the OpenAI format and get standard tool_calls back; json_schema structured output works too.
Do repeated requests hit the cache?
Implicit caching is supported; explicit caches are not. Repeated long inputs produced hits in our tests, billed at one twentieth of the input rate ($0.0225 per million tokens). A hit on every repeat is not guaranteed.
What does Beta mean?
During Beta the upstream capacity can vary. Failed calls are never billed and can be resubmitted as they are. We drop the Beta label once features and prices are stable.

Related Models

Explore other models you can integrate.

View all →