The model's exact ID on OpenRouter is tencent/hy4-flash, and it can be called through OpenRouter's Chat Completions API.
Suitable tasks: Calling Hy4 Flash through the OpenRouter API; the model card positions it for coding assistance, agent workflows, document and chart understanding, and high-frequency routine automation.
Unsuitable tasks: Tasks that require routing through OpenRouter's EU region; the model card indicates that this route is not currently supported for this model.
Applicable model version: Tencent Hy4 Flash, model ID tencent/hy4-flash.
Applicable client, agent, or API: OpenRouter Chat Completions API; the panel also lists endpoints for the Responses API and Anthropic Messages API.
Recommended reasoning tier and parameters: The model card does not provide a recommended tier. It only lists available parameters; no parameter values are inferred here.
First, create an API key on OpenRouter and set it as an environment variable:
export OPENROUTER_API_KEY="<OPENROUTER_API_KEY>"The following is the streaming Chat Completions configuration shown in the model card's API panel:
curl -N https://openrouter.ai/api/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $OPENROUTER_API_KEY" \
-d '{
"model": "tencent/hy4-flash",
"stream": true,
"messages": [
{"role": "user", "content": "Hello"}
]
}'The panel lists HTTP-Referer (site URL) and X-Title (site name) as optional OpenRouter-specific request headers. They are used to display the app on OpenRouter's leaderboard; the example does not require them.
Create an API key on OpenRouter and set OPENROUTER_API_KEY.
Send a POST request to https://openrouter.ai/api/v1/chat/completions.
Set model to tencent/hy4-flash and pass the conversation content in messages.
To receive a streaming response, set stream: true; the model card says the endpoint supports both streaming and non-streaming modes.
The API panel lists the following parameters and descriptions:
| Parameter | Type | Default | Description on the page |
|---|---|---|---|
reasoning | map | Not specified | Controls reasoning behavior for models that support thinking tokens. This includes whether reasoning is enabled, its effort level, the maximum number of reasoning tokens, and whether reasoning content is excluded from the response. |
response_format | map | Not specified | Forces a specific output format. The FAQ says structured output via JSON Schema is supported. |
tools | array | Not specified | Tool-calling parameters, using the OpenAI tool-calling request format. |
tool_choice | string or object | Not specified | Controls whether and how tools are called. The FAQ confirms that tool calling is supported. |
temperature | float | 1 | Affects response diversity. |
stop | array | Not specified | Stops generation as soon as a token specified in the array is encountered. |
max_completion_tokens | integer | Not specified | Sets the maximum number of tokens in the generated response. |
reasoning_effort | enum | Not specified | An OpenAI-style setting for reasoning effort. The page does not list the available enum values. |
max_tokens | integer | Not specified | Sets the maximum number of tokens in the generated response. |
Both the model card title and the API panel show the model ID as tencent/hy4-flash.
The API panel gives https://openrouter.ai/api/v1/chat/completions as the Chat Completions URL, with POST as the method and Bearer Key as the authentication method.
The panel also lists https://openrouter.ai/api/v1/responses and https://openrouter.ai/api/v1/messages, but this configuration example uses Chat Completions.
The FAQ lists a context window of 1,048,576 tokens and a maximum completion of 65,536 tokens; inputs are text and images, and outputs are text.
The FAQ says tools and tool_choice are supported, and JSON Schema can be used in response_format.
The page shows a model release date of 2026-11-16 and says the model is hosted by Tencent Cloud as its sole provider on OpenRouter.
At the time of collection, the page listed input pricing of $0.035 per million tokens, output pricing of $0.1405 per million tokens, and cache-read pricing of $0.0015 per million tokens. It also showed a 50% discount. Prices may change; check the model card's current pricing before use.
The parameter table lists parameters displayed in the OpenRouter panel. The page does not specify per-parameter constraints for this model or give enum options for reasoning_effort, so specific supported values cannot be inferred from it.
OpenRouter's API panel says the reasoning parameter applies to models that support thinking tokens; the page does not detail Hy4 Flash-specific reasoning configuration in this parameter table.
The sole-provider information means that Tencent Cloud was the only provider listed when the model card was collected; providers and prices may change.
The page says EU-region routing is currently unsupported. This records only the status shown on the OpenRouter model card on the collection date.
The tasks described by the model card are the platform's characterization, not independent test results.
Create your own API key on OpenRouter, save it in an environment variable, and then run the cURL command above. The example requests a streaming text response from the model; the actual response, availability, and price depend on OpenRouter's service status at the time of the request. The model card and API panel were viewable without logging in at the time of collection.
Hy4 Flash