# Tencent Cloud TokenHub (slim)

> Tencent Cloud TokenHub is a hosted model gateway serving Tencent's Hy models and third-party models such as DeepSeek, Kimi, GLM and MiniMax through OpenAI-compatible and Anthropic-compatible endpoints in Guangzhou and Singapore, with per-key scopes and quotas.

- Full: https://www.anchorterminal.com/tools/tencent-tokenhub.md (~11,200 tokens) · this version ~2,480 tokens · JSON https://www.anchorterminal.com/tools/tencent-tokenhub.json · canonical https://www.anchorterminal.com/tools/tencent-tokenhub
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-11-02 · pricing https://www.anchorterminal.com/tools/tencent-tokenhub#pricing · reviews https://www.anchorterminal.com/tools/tencent-tokenhub#reviews
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

**C · 59.5/100 · rank #646 of 1260 · #18 in Model APIs & inference · not agent-ready · confidence medium**


Assessment: Keys can be limited to chosen models, source IPs and token quotas, errors name their source and code, and a 99.5% SLA covers paid use of models Tencent runs. Sign-up needs a Tencent Cloud account with real-name verification, prices are in CNY, no public status page was found, and 18 model retirement notices ran from April to October 2026.

## Facts

- Kind: Model router · vendor: Tencent Cloud · category: Model APIs & inference · legal entity: Tencent Cloud Computing (Beijing) Co., Ltd. (腾讯云计算（北京）有限责任公司) · provenance 49/100
- Endpoint: `https://tokenhub.tencentmaas.com/v1` (HTTP)
- Auth: API key · pricing: Pay per use · x402: no · licence: Proprietary service under the TokenHub Terms of Service, the TokenHub SLA and the Tencent Cloud Service Agreement, with a special note for third-party deployed models. The `tencent-tokenhub-cli` package is Apache-2.0
- Probe metrics: not measured yet (probes haven't run)
- Endpoints: Guangzhou https://tokenhub.tencentmaas.com (mainland China resources) and Singapore https://tokenhub-intl.tencentmaas.com (global), with backup hosts on tencentmaas.cn. Paths `/v1/chat/completions`, `/v1/responses`, `/v1/messages` and `/v1/models`. Calls across regions or sites are not supported
- Models on 1 November 2026: The API usage page lists Hy4 preview, Hy3, Hy-MT2 translation and Hy-Role models from Tencent, DeepSeek V4 and V4.1 (some supplied directly by DeepSeek), GLM-5.2 and 5.3, Kimi K2.6 to K3.1 Pro, MiniMax M2.7 to M3.1, Qwen3.5, MiMo and Step 5 Preview, plus image, video, 3D, speech, music, vision and embedding models
- Prices: CNY per 1M tokens, the same in Guangzhou and Singapore for most models. Hy4 preview 6 in, 18 out, 0.3 cached. Hy3 1, 4, 0.25. DeepSeek-V4.1-Flash 1, 4, 0.02 off-peak and 2, 8, 0.04 at weekday peak. Kimi K3.1 Pro 13, 68, 1.3. GLM-5.3 8, 28, 2 in Guangzhou. TokenHub web search 7 or 12 CNY per 1,000 calls
- Free allowance: One free pack per account, 1M tokens per language and multimodal model and credits for some video and 3D models, valid one year from claiming. The campaign runs to 31 December 2026. Calls stop when it runs out unless pay-as-you-go is switched on
- Subscriptions: Token Plan for individuals (Hy Token Plan from 28 to 468 CNY a month, the general plan from 39 to 599 CNY), Token Plan for enterprises, Coding Plan, reserved TPM and dedicated model units through a sales manager
- Rate limits: Per model and per account, shown in the console and returned by DescribeModelQuota. Increases are self-serve when weekly peak use is above 60%, at most twice a week. Batch jobs 20 created a minute and 10 active per account. Image models 5 concurrent
- Errors: JSON `error` with `code` (six digits), `message`, `message_zh`, `type`, `source` (client, gateway or upstream), `upstream_code`, `upstream_status` and `request_id`. 30 codes documented, including 402 for an exhausted free pack, 451 for filtered content and six 429 codes. Stream errors arrive as SSE error frames
- Credentials: Long-lived API keys, 50 per account by default. Scope to all or chosen models and inference services, IP allowlist of up to 200 ranges, daily, monthly or total token quota, enable, disable and delete. Control-plane calls (CreateApiKey and others) use Tencent Cloud API 3.0 signatures
- Batch: Hy3 and Hy4 preview at half the online price. JSONL in the customer's own COS bucket, one model per file, completion window set in hours, results written back to COS
- Data use in the terms: Clause 2.17 says inputs are not used to train Tencent Cloud's own models. Clause 2.13 places processing generally in mainland China for the China site and in Singapore for the international site. Clause 2.14 keeps data only as long as the service needs, with no periods stated
- Logs and monitoring: Audit logs and full call logs can be delivered to the customer's CLS (billed separately). The console shows usage by key, model and service, and TTFT, TPOT, error and 429 rates
- Optional security: LLM-WAF checks inputs and outputs for prompt injection, banned content and sensitive data, in Guangzhou only and billed separately. PrivateLink connections from a VPC keep calls off the public internet
- Contract: Tencent Cloud Computing (Beijing) Co., Ltd. (腾讯云计算（北京）有限责任公司) under the TokenHub Terms of Service (effective 14 September 2026) and SLA (effective 23 April 2026), supplementing the Tencent Cloud Service Agreement
- Support: Tickets and phone support around the clock, presales chat and the Tencent Cloud Q&A community
- 2026-09-30 Shutdown: Tencent Cloud's older model platforms (Hunyuan, the knowledge engine's DeepSeek API) shut, with users moved to TokenHub
- 2026-10-09 Shutdown: GLM-5, GLM-5-Turbo and GLM-5.1 retired, with callers switched to GLM-5.3 and GLM-5.1 batch inference ended without migration
- 2026-10-17 Notice: DeepSeek-V4.1-Flash (directly supplied) switched to the DeepSeek-V4.1-Flash-1017 version under the same ids, announced on 19 October
- 2026-10-31 Shutdown: DeepSeek-V4-Flash 0731 retired, with eligible callers switched to the newest DeepSeek model
- 2026-11-01 Price change: Web search Search and Fetch calls of DeepSeek's directly supplied models charged at 5 CNY per 1,000 successful requests
- 2026-11-12 Shutdown: Hy-Image-3.0 retired, with Hy-Image-3.5-preview recommended
- 2026-11-17 Shutdown: MiniMax-M2.7 retired, with eligible callers switched to MiniMax-M3
- 2026-11-24 Shutdown: Seedream-Image-v5.0-lite retired, with Seedream-Image-v5.0-flash recommended

- Scores: Reliability 46, Performance pending, Schema & documentation 73, Agent ergonomics 83, Security & auth 71, Payments & pricing 35, Task success pending, Maintenance & community 73, Transparency & trust 54 · negative events -3 · total over the 7 assessed categories
- Why: Reliability, Hosted reading. · Schema & documentation, Router, scored on the API lines. · Agent ergonomics, Router, read on the model lines. · Security & auth, Router, read on the model lines. · Payments & pricing, No x402, MPP or L402 (0). · Maintenance & community, Closed service, read on the model lines. · Transparency & trust, Closed service under the TokenHub Terms of Service (effective 14 September 2026), the SLA (23 April 2026), a special note on third-party dep…
- Sources: 49, open questions: 7, both in the full twin
- Capabilities: inference.router, inference.llm, inference.open-weights, embed.text, embed.multimodal, image.generate, video.generate, speech.tts, speech.stt, audio.music, translate.text, web.search, obs.gateway
- JSON: https://www.anchorterminal.com/api/v1/tools/tencent-tokenhub.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/tencent-tokenhub.svg` or a link to https://www.anchorterminal.com/tools/tencent-tokenhub from a page on tencentmaas.com or cloud.tencent.com or one of their subdomains, or the README at the root of github.com/TencentCloud/THCLI, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.
- Support: https://cloud.tencent.com/document/product/1823/130109; https://console.cloud.tencent.com/workorder/category; https://cloud.tencent.com/act/event/Online_service; more in the full twin

## Before you call it

1. Use `https://tokenhub.tencentmaas.com/v1` with keys made for Guangzhou and `https://tokenhub-intl.tencentmaas.com/v1` with keys made for Singapore. Keys do not work across the two
2. Send the key as `Authorization: Bearer` on the OpenAI paths and as `x-api-key` on `/v1/messages`
3. Parse `error.code` as a string or a number, since 429 codes can come back as integers, and wait for `Retry-After` before retrying
4. Take model ids from `GET /v1/models` and watch for `pre-offline`. The quickstart's `deepseek-v3` example is not in the current model list
5. Expect DeepSeek prices to double on weekdays from 9am to noon and 2pm to 6pm Beijing time, when the peak rate applies

## Connect

```bash
npm i -g tencent-tokenhub-cli
```

```bash
curl https://tokenhub.tencentmaas.com/v1/models \
  -H "Authorization: Bearer $TOKENHUB_API_KEY"
```

```bash
{
  "env": {
    "ANTHROPIC_BASE_URL": "https://tokenhub.tencentmaas.com",
    "ANTHROPIC_AUTH_TOKEN": "YOUR_API_KEY",
    "ANTHROPIC_MODEL": "hy3"
  }
}
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| SiliconFlow | D | 46.8 | inference.llm, inference.open-weights, embed.text, image.generate, video.generate, speech.tts, speech.stt | https://www.anchorterminal.com/tools/siliconflow.min.md |
| Cloudflare Workers AI | B | 69.2 | inference.llm, inference.open-weights, embed.text, image.generate, speech.stt, speech.tts | https://www.anchorterminal.com/tools/cloudflare-workers-ai.min.md |
| Vercel AI Gateway | B | 66.8 | inference.router, inference.llm, obs.gateway, embed.text, image.generate, video.generate | https://www.anchorterminal.com/tools/vercel-ai-gateway.min.md |
| DeepInfra | B | 63.1 | inference.llm, inference.open-weights, embed.text, image.generate, speech.stt, speech.tts | https://www.anchorterminal.com/tools/deepinfra.min.md |
| LocalAI | B | 62.5 | inference.open-weights, embed.text, speech.stt, speech.tts, image.generate, video.generate | https://www.anchorterminal.com/tools/localai.min.md |


## Panel reviews (0, desk reviews from public material, no calls made)