When calling the model through OpenRouter, use the model ID google/gemini-4-hydrogen. OpenRouter first selects an upstream provider according to the routing mode. Standard, Flex, and Priority are endpoint types listed on the page; pricing, latency, and availability vary by provider and over time.
Suitable tasks: Connecting to Gemini 4 Hydrogen through a single OpenRouter API and viewing or adjusting request routing.
Unsuitable tasks: Configurations that require a fixed Google upstream endpoint or pricing that remains unchanged over time.
Applicable model version: google/gemini-4-hydrogen.
Applicable clients, agents, or APIs: OpenRouter Chat Completions API; the page also shows examples using the OpenRouter SDK.
Recommended reasoning levels and parameters: The page lists the low, medium, high, and xhigh reasoning levels. It does not specify their API field syntax, so no parameters are inferred here.
The request format and model ID below come from the cURL example shown on OpenRouter's page. Set the OPENROUTER_API_KEY environment variable to your own OpenRouter API key, then send the request.
curl -N https://openrouter.ai/api/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $OPENROUTER_API_KEY" \
-d '{
"model": "google/gemini-4-hydrogen",
"stream": true,
"messages": [
{"role": "user", "content": "Hello"}
]
}'Confirm on OpenRouter's model page that the model ID is google/gemini-4-hydrogen.
Make a call using an OpenRouter API key and the request format above.
Choose a routing strategy for the task: the page describes Balanced as balancing price and speed, Nitro as the fastest, and Floor as the lowest-priced option.
To compare upstream providers, check the provider information for Standard, Flex, and Priority endpoints and their pricing and performance data. Endpoints marked “Not used in Standard routing” are not part of Standard routing.
The page says OpenRouter will try switching to another healthy provider when an upstream error occurs. If the request's provider filters do not allow other endpoints, the switch may fail.
As of 2026-11-16, the model page listed the following information. The page data is dynamic and reflects only what was visible at the time of collection:
| Page label | Provider / endpoint | Input / output price (per million tokens) | Other information shown on the page |
|---|---|---|---|
| Standard | Google Vertex | $1.50 / $7.50 | Latency: 3.83 seconds; throughput: 105 tok/s; uptime: 99.80% |
| Standard | Google AI Studio | $1.50 / $7.50 | Latency: 2.98 seconds; throughput: 174 tok/s; uptime: 96.14% |
| Non-Standard | Google AI Studio Flex | $0.75 / $3.75 | Latency: 2.12 seconds; throughput: 163 tok/s; uptime: 100.00% |
| Non-Standard | Google Vertex | $1.65 / $8.25 | Latency and throughput not shown; uptime: 100.00% |
| Non-Standard | Google Vertex | $1.65 / $8.25 | Latency: 1.56 seconds; throughput: 126 tok/s; uptime: 100.00% |
| Non-Standard | Google Vertex Priority | $2.70 / $13.50 | Latency: 2.02 seconds; throughput: 141 tok/s; uptime: 100.00% |
| Non-Standard | Google AI Studio Priority | $2.70 / $13.50 | Latency and throughput not shown; uptime: 100.00% |
The page overview lists text, images, audio, video, and files as supported input types, and specifies a 1.0M context window. Its model description says continuation can support up to 1M output tokens. Verify these capabilities and prices against the model page and API response at the time of the request.
OpenRouter uses google/gemini-4-hydrogen as its platform model ID. This identifier is for OpenRouter requests and does not imply the model ID used by other APIs.
Upstream providers, routing strategies, endpoint availability, pricing, latency, throughput, and uptime may change. The page's pricing and performance figures are observations from 2026-11-16, not current quotes or service guarantees.
The model page does not show an example request field for pinning a specific upstream provider. To pin a provider, configure the request using OpenRouter's current routing documentation and check the result; do not infer the parameter name from this page alone.
The page lists reasoning levels, but this note does not infer request fields or parameter compatibility that it does not show.
During collection, the original URL was opened in Tabbit and the model page and API example were checked. The API view on the page can be used to verify the model ID and cURL request. Pricing and routing data change as the page is updated, so retrieve them again on the day of reproduction and record the date.
Gemini 4 Hydrogen