OpenClaw's Tencent plugin provides TokenHub and TokenPlan endpoints. The Hy4-Flash model references are tencent-tokenhub/hy4-flash and tencent-tokenplan/hy4-flash, respectively.
Suitable tasks: Configure Tencent Cloud TokenHub or TokenPlan in OpenClaw and select Hy4-Flash for the chosen endpoint.
Unsuitable tasks: Use this article's provider IDs, authentication variables, or model references with other clients or APIs.
Applicable model version: Tencent Hunyuan Hy4-Flash.
Applicable client, agent, or API: OpenClaw's official Tencent provider plugin, @openclaw/tencent-provider.
Recommended reasoning level and parameters: high by default. OpenClaw sends reasoning_effort as none or high; all other reasoning levels map to high.
| Endpoint | Model reference | Credential environment variable | OpenClaw auth option |
|---|---|---|---|
| TokenHub | tencent-tokenhub/hy4-flash | TOKENHUB_API_KEY | tokenhub-api-key |
| TokenPlan | tencent-tokenplan/hy4-flash | TOKENPLAN_API_KEY | tokenplan-api-key |
Both endpoint catalogs mark Hy4-Flash as a fast-tier model that supports image input and reasoning. Its context window is 1,048,576 tokens, and its maximum output is 65,536 tokens.
First, create an API key in the corresponding Tencent Cloud service and place it in the current shell environment. Run the following commands on the Gateway host; --accept-risk is required in non-interactive mode.
TokenHub:
export TOKENHUB_API_KEY="<TOKENHUB_API_KEY>"
openclaw onboard --non-interactive \
--mode local \
--auth-choice tokenhub-api-key \
--tokenhub-api-key "$TOKENHUB_API_KEY" \
--skip-health \
--accept-riskTokenPlan:
export TOKENPLAN_API_KEY="<TOKENPLAN_API_KEY>"
openclaw onboard --non-interactive \
--mode local \
--auth-choice tokenplan-api-key \
--tokenplan-api-key "$TOKENPLAN_API_KEY" \
--skip-health \
--accept-riskList the model catalogs for both endpoints to confirm that Hy4-Flash is available:
openclaw models list --provider tencent-tokenhub
openclaw models list --provider tencent-tokenplanWhen selecting a model in OpenClaw's configuration, use the full model reference from the table above that corresponds to the chosen endpoint. The source documentation does not specify the configuration key or command for setting a default model.
reasoning_effort input | Value OpenClaw sends to Tencent |
|---|---|
none | none |
high | high |
| Other reasoning levels | high |
| Not specified | high by default |
Choose TokenHub or TokenPlan and create an API key for that endpoint.
Run the corresponding non-interactive onboarding command on the host running the Gateway.
Run the corresponding openclaw models list --provider ... command and check that hy4-flash appears in the catalog.
Set the model configuration reference to the full model ID for the corresponding endpoint. Source note: new onboarding selects hy4-preview by default; it does not automatically switch to Hy4-Flash.
If you need to adjust reasoning strength, use only the mappings explicitly described in the source; if unspecified, treat it as high.
The official provider page lists provider IDs tencent-tokenhub and tencent-tokenplan, the plugin package @openclaw/tencent-provider, and the API type openai-completions, which is OpenAI-compatible.
The TokenHub API key environment variable is TOKENHUB_API_KEY, its auth option is tokenhub-api-key, and its direct CLI argument is --tokenhub-api-key <key>. The corresponding TokenPlan values are TOKENPLAN_API_KEY, tokenplan-api-key, and --tokenplan-api-key <key>.
The page's built-in model catalogs list tencent-tokenhub/hy4-flash and tencent-tokenplan/hy4-flash, both described as supporting images and reasoning and belonging to the fast tier.
The page says Hy4-Flash accepts the same image formats as Hy4, has an input limit of 983,040 tokens and a context window of 1,048,576 tokens; the catalogs list a maximum output of 65,536 tokens.
OpenClaw's reasoning levels for Hy4-Flash are none or high, with high as the default; all other reasoning levels map to high.
The default TokenHub base URL is https://tokenhub.tencentmaas.com/v1, and its international override is https://tokenhub-intl.tencentmaas.com/v1. The default TokenPlan base URL is https://api.lkeap.cloud.tencent.com/plan/v3.
New onboarding selects hy4-preview by default for both endpoints. The page says models can be listed in the OpenClaw catalog or called using a manually entered or configured model reference.
The selected API key must have access to the corresponding model. Restricted keys must include the model you plan to use in their authorization scope.
Remote-mode onboarding (--mode remote) configures only the local client's connection; it does not configure provider credentials on the server. Set credentials on the Gateway host.
The TokenHub international URL is an override listed in the documentation, not the default endpoint.
The source page describes OpenClaw plugin behavior. It does not provide the specific configuration key for setting a default model or explain whether reasoning-level mappings differ between TokenHub and TokenPlan, so this article does not infer those configurations.
The page describes Hy4-Flash as fast tier; this is not an independent performance test. This article does not infer specific latency or quality advantages from that description.
Create an API key for the corresponding Tencent Cloud endpoint, put it in the matching environment variable, run the onboarding command on the Gateway host, and then use the model catalog command to verify the provider model. Do not put a real API key in configuration documentation or command history. The catalog and parameters recorded here were visible on the official OpenClaw provider page when collected on 2026-11-17.
Hy4 Flash