How to Deploy the Open‑Source QwQ‑32B Inference Model on Alibaba Cloud CAP

QwQ‑32B, a newly released open‑source inference model, quickly gained global attention for its strong performance across various scenarios.

Powered by Alibaba Cloud Function Compute (FC), the Serverless + AI cloud‑native development platform CAP now offers both model service and application‑template deployment options, allowing you to launch QwQ‑32B via a one‑click template or API calls.

Benchmark Highlights

In tests such as the AIME24 math set, LiveCodeBench coding evaluation, LiveBench (led by Meta’s chief scientist), IFEval instruction‑following, and the BFCL function‑calling benchmark, QwQ‑32B matches or surpasses leading models like DeepSeek‑R1‑671B, DeepSeek‑R1‑Distilled‑Qwen‑32B, OpenAI‑o1‑mini, and others.

Preparation

1. Open the CAP console, authorize access, and wait for the authorization to complete.

2. The GPU function created in Function Compute is billed by resource‑time; unused functions only incur snapshot fees. Claim the Function Compute trial quota to offset costs.

Method One: Application‑Template Deployment

Step 1 – Create Project

Enter the CAP console and click “Create based on template”.

Step 2 – Deploy Template

Search for “QWQ” and select the “Qwen‑QwQ inference model chat assistant” template, then click “Deploy Now”.

Choose a region (Beijing, Shanghai, or Hangzhou), review the billing preview, and confirm deployment (approximately 10 minutes).

Notes

Select a region close to your resources; if using a NAS file system, match the region.

If GPU resources are insufficient, switch regions and retry.

Method Two: Model‑Service Deployment (API)

Step 1 – Create Blank Project

In the CAP console, click “Create blank project” and name it.

Step 2 – Choose Model Service

Select the QwQ‑32B‑GGUF model (currently only available in Hangzhou).

Step 3 – Deploy Model Service

Configure resources (Ada series is recommended) or customize GPU type and specs.

Preview billing, then confirm deployment (model download takes 10‑30 minutes).

Step 4 – Verify Model Service

Click “Debug” to test the model, use the Open‑WebUI for interactive chat, or invoke the model from the command line.

You can also test via third‑party platforms such as Chatbox.

Project Deletion

Open project details and click “Delete”.

Confirm which services to retain or remove.

Check the acknowledgment box and confirm deletion.

Reference Links

1. https://www.aliyun.com/product/cap

2. https://common-buy.aliyun.com/package

3. https://help.aliyun.com/zh/functioncompute/fc-3-0/product-overview/billing-overview-1

4. https://cap.console.aliyun.com/projects

5. https://help.aliyun.com/zh/cap/product-overview/billing-overview

6. https://cap.console.aliyun.com/projects

7. https://help.aliyun.com/zh/cap/product-overview/billing-overview

8. https://web.chatboxai.app/