Deploy Qwen2.5 LLM on Alibaba Cloud Function Compute: A Step‑by‑Step Guide

Solution Overview

Qwen2.5 is a large‑scale language and multimodal model that excels at long‑text processing, integrates extensive domain knowledge, supports more than 29 languages, and provides strong coding and mathematics capabilities.

Deploying Qwen2.5 on Alibaba Cloud Function Compute (FC)

Function Compute offers a serverless environment with automatic scaling and flexible GPU billing (pay‑as‑you‑go, tiered, ultra‑fast). By deploying Qwen2.5 through Ollama and exposing it via Open WebUI, users can handle high‑concurrency inference while only paying for active resources.

Step 1 – Deploy the Ollama Application

Open the Ollama template URL and create a new application.

https://fcnext.console.aliyun.com/applications/create?template=ollama-qwen2_5&deployType=template-direct&from=solution

Select the desired Qwen2.5 model size (1.5B, 3B, or 7B) from the Model Name dropdown.

Leave all other configuration items at their defaults and click Create and Deploy Default Environment . The application is deployed as shown below.

Step 2 – Deploy the Open WebUI Application

Open the Open WebUI template URL and create a new application.

https://fcnext.console.aliyun.com/applications/create?template=fc-open-webui&deployType=template-direct

In Advanced Configuration > Region , select the same region used for the Ollama app.

Enable authentication for production use.

Enter the internal HTTP trigger address of the Ollama service (obtained in Step 3) as the Open WebUI endpoint.

Click Create and Deploy Default Environment . After deployment, open the provided domain to access Open WebUI.

Warning: The region selected for Open WebUI must match the region of the Ollama application.

Step 3 – Retrieve the Ollama Internal HTTP Trigger Address

Open the Function Compute console, locate the Ollama application, and open its detail page.

In the Function Resources section, click the function name to view details.

Hover over the HTTP Trigger entry and copy the displayed internal URL.

Using Open WebUI to Call Qwen2.5

After logging into Open WebUI, select Select a model and choose the deployed Qwen2.5 model. You can then interact with the model via the chat box.

Example capabilities:

Multilingual response (e.g., self‑introduction in French).

Enhanced coding and mathematics reasoning thanks to domain‑expert fine‑tuning.

Document summarization: upload a local file (e.g., 百炼手机详细参数.docx ) and prompt “Summarize document content”. The model extracts key information.

Configuring Open WebUI Language to Simplified Chinese

Click the settings icon (top‑right) and choose Settings .

Navigate to General > Language and select Chinese（简体中文） .

Save the changes; the interface refreshes in Chinese.

Automatic Scaling of Function Compute

FC automatically scales the number of Ollama function instances based on request volume. When traffic spikes, new instances are created; idle instances are terminated after 3–5 minutes, reducing cost while maintaining low latency.

Related Links

Ollama template URL:

https://fcnext.console.aliyun.com/applications/create?template=ollama-qwen2_5&deployType=template-direct&from=solution

Open WebUI template URL:

https://fcnext.console.aliyun.com/applications/create?template=fc-open-webui&deployType=template-direct

Function Compute console: https://fcnext.console.aliyun.com/applications Sample document for summarization:

https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20240701/geijms/%E7%99%BE%E7%82%BC%E7%B3%BB%E5%88%97%E6%89%8B%E6%9C%BA%E4%BA%A7%E5%93%81%E4%BB%8B%E7%BB%8D.docx