Tencent Cloud TokenHub uses hy4-flash as the model parameter value. Its Chat Completions endpoint uses the Tencent Cloud TokenHub URL and a Bearer API key. Deep thinking is enabled by default, with the default reasoning effort set to high.
Suitable tasks: Calling Hy4-Flash through Tencent Cloud TokenHub and explicitly setting the thinking toggle or reasoning effort.
Unsuitable tasks: Using these parameters with OpenRouter or another provider's API.
Applicable model version: TokenHub model parameter value hy4-flash.
Applicable client, agent, or API: The documentation confirms compatibility with OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages. The example below uses Chat Completions.
Recommended reasoning effort and parameters: high by default; the model supports high and no_think.
The following is a minimal Chat Completions request with hy4-flash filled in according to TokenHub's official model parameter table. Replace YOUR_API_KEY with an API key created in the TokenHub console:
curl -X POST 'https://tokenhub.tencentmaas.com/v1/chat/completions' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer YOUR_API_KEY' \
-d '{
"model": "hy4-flash",
"messages": [
{"role": "user", "content": "你好"}
],
"stream": false
}'To explicitly set the high reasoning effort, add this to the request JSON:
"reasoning_effort": "high"reasoning_effort can also be set to no_think. The documentation also lists a thinking toggle: {"type":"enabled"} turns it on, and {"type":"disabled"} turns it off. Hy4-Flash has thinking enabled by default. The documentation does not specify which setting takes precedence when thinking and reasoning_effort are both set.
Enable TokenHub in Tencent Cloud and create an API key.
Send a POST request to https://tokenhub.tencentmaas.com/v1/chat/completions.
Authenticate with Authorization: Bearer YOUR_API_KEY, set model to hy4-flash, and provide messages.
Use the default high reasoning effort. To change the reasoning mode, set reasoning_effort or thinking as needed.
The Hunyuan API guide lists hy4-flash as a model parameter value and notes compatibility with the OpenAI Chat Completions API, OpenAI Responses API, and Anthropic Messages API.
The same document lists a 1024k context window for Hy4-Flash, with a maximum input of 960k and maximum output of 64k.
Tencent Cloud TokenHub's deep thinking documentation lists hy4-flash as the model parameter value and enabled as the default thinking state. The default reasoning depth is high, and supported values are high and no_think.
The generic cURL example in the Hunyuan API guide uses hy4 as the model value. This article substitutes the exact model value hy4-flash, based on that document's Hy4-Flash model table and the parameter table in the deep thinking documentation; it is not a verbatim example from the source.
See Tencent Cloud TokenHub deep thinking documentation for the reasoning parameters. That page is a supplementary source for parameter definitions; the primary source for this article remains the Hunyuan API guide cited above.
1024k, 960k, and 64k are the documented limits for context, maximum input, and maximum output. The documentation does not specify a more detailed unit definition or additional request-side limits here.
The compatibility statement covers three protocols, but this article only provides a Chat Completions request example. Check the documentation for each protocol for the specific paths and fields used by Responses and Anthropic Messages.
Do not apply OpenRouter model IDs, endpoint paths, prices, or default parameter values to TokenHub.
Create an API key in the TokenHub console, replace the placeholder in the example, and run the cURL command. The request is sent to Tencent Cloud TokenHub; the example does not include a secret. The Hunyuan API guide and deep thinking documentation were last updated on 2026-11-16. Service capabilities and parameters may change as the official documentation is updated.
Hy4 Flash