The model ID for the production version of Hy4 on OpenRouter is tencent/hy4, and it is called through the OpenRouter API. The model page lists Tencent Cloud as the only provider and states that OpenRouter forwards requests directly without further provider routing.
Suitable tasks: Calling Hy4 through OpenRouter's compatible API, with support for configuring reasoning, streaming responses, or tool calls.
Unsuitable tasks: Treating tencent/hy4 as the model ID for Tencent Cloud TokenHub; assuming OpenRouter provider prices, discounts, or routing status will remain unchanged over time.
Applicable model version: Hy4 as listed on the OpenRouter model page, marked as released on 2026-11-04; model ID tencent/hy4.
Applicable clients, agents, or APIs: OpenRouter Chat Completions, Responses, and Anthropic Messages APIs. The example uses the OpenRouter TypeScript SDK and Chat Completions.
Recommended reasoning tier and parameters: The page lists a reasoning object and the OpenAI-style reasoning_effort parameter, but gives no recommended or default reasoning tier. Set these according to the API documentation and client support.
The following cURL example is based on the Chat Completions endpoint and model ID shown in the OpenRouter API panel. First, put your API key in an environment variable:
export OPENROUTER_API_KEY="YOUR_OPENROUTER_API_KEY"
curl -N https://openrouter.ai/api/v1/chat/completions \
-H "Authorization: Bearer $OPENROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "tencent/hy4",
"stream": true,
"messages": [
{"role": "user", "content": "请用三句话概括这段文字。"}
]
}'Chat Completions: POST https://openrouter.ai/api/v1/chat/completions
Responses: POST https://openrouter.ai/api/v1/responses
Anthropic Messages: POST https://openrouter.ai/api/v1/messages
The API panel shows the model value tencent/hy4 for all three interfaces. Authentication uses Authorization: Bearer $OPENROUTER_API_KEY, and the request type is application/json.
HTTP-Referer and X-Title are optional headers. The panel says that setting them can list the application in the OpenRouter rankings.
Parameters listed in the API panel: reasoning, temperature (default 1), stop, max_completion_tokens, reasoning_effort, max_tokens, response_format, tools, and tool_choice. The page says the reasoning object can control whether reasoning is enabled, reasoning effort, the maximum number of reasoning tokens, and whether reasoning content is excluded from the response.
OpenRouter says responses from reasoning models may include reasoning_details; for multi-turn conversations, preserve the complete reasoning_details in the messages sent back. Do not confuse this with the reasoning_content field in the TokenHub documentation.
Set the key in the OpenRouter API key environment variable.
Set the endpoint for the chosen interface and use model: "tencent/hy4".
Set stream: true when a streaming response is needed; the Chat Completions example uses SSE.
To control reasoning, set reasoning or reasoning_effort according to OpenRouter's parameter documentation. For multi-turn requests, preserve the complete reasoning_details from the response.
Check the model ID, finish status, and usage in the response to confirm the call result.
The model page is titled “Tencent: Hy4,” describes it as the production version of Hy4, and lists its release date as 2026-11-04 and context window as 1.0M.
The examples in the API panel and all three interface configurations use the model ID tencent/hy4.
The provider list contains only Tencent Cloud. The page says the model is hosted by one provider and that OpenRouter forwards each request directly without further routing.
Price snapshot (as shown on the page on 2026-11-16): listed input price $0.834 per million tokens and output price $2.501 per million tokens; the same page shows 50% off, with displayed prices of $0.417 per million tokens for input and $1.251 per million tokens for output. The listed cache read price is $0.042 per million tokens, discounted to $0.021 per million tokens.
The OpenRouter API panel provides example tabs for the TypeScript SDK, Python, TypeScript fetch, cURL, Python OpenAI SDK, and TypeScript OpenAI SDK. The readable cURL example calls /api/v1/chat/completions and uses tencent/hy4.
The OpenRouter page shows platform-side model IDs, API entry points, routing information, and a price snapshot; it does not replace Tencent Cloud TokenHub's model ID and parameter documentation. TokenHub uses hy4; OpenRouter uses tencent/hy4.
Although the page lists multiple compatible interfaces and general parameters, the model's API panel does not specify Hy4's actual support for each parameter, nor does it provide a recommended reasoning_effort value. Before deployment, check the current documentation for the selected interface and SDK.
The provider, routing, and discounted prices shown on the page are a snapshot from 2026-11-16. Discounts, final charges, and provider status may change; use the OpenRouter page and bill at the time of the call as the source of truth.
The example shows a regular text conversation and a streaming request. This article does not infer image or PDF input limits from it, and tool-call behavior was not tested.
On 2026-11-16, Tabbit was used to open the model page at https://openrouter.ai/tencent/hy4 and inspect its API panel, interface endpoints, parameter table, provider information, context window, and displayed prices. This public information was accessible without signing in. No model request was sent because an API key was not configured, so actual responses and charges were not verified.
Hy4