Skip to main content
Qwen Cloud home page
Docs
Clients & tools

Cherry Studio

Desktop LLM client with MCP

Cherry Studio is a desktop AI client that combines chat, MCP (Model Context Protocol) tools, and local knowledge bases. Connect it directly to Qwen Cloud's pay-as-you-go API endpoints for powerful AI capabilities with external tool integration.

Quick start

Get running in a few minutes: 
# 1. Install
Download Cherry Studio from cherry-ai.com

# 2. Configure (open Settings > find Model Provider > select Alibaba Cloud)
API Key: sk-xxx
API Address: https://dashscope-intl.aliyuncs.com/compatible-mode/v1
Click "Add" button

# 3. Test (add a model and chat)
In Model ID field, type: qwen3.5-plus
Click to add the model
Send a message: "Explain how neural networks work"
You should see: qwen3.5-plus explains neural networks with detailed information

Configuration

Basic setup

Configure Cherry Studio to use Qwen Cloud:
  • API endpoint: https://dashscope-intl.aliyuncs.com/compatible-mode/v1
  • Authentication: API key required
  • Model selection: Enter any Qwen model ID

Step-by-step configuration

1

Open settings

Click the settings button in the upper-right corner
2

Find Alibaba Cloud

In Model Provider list, find Alibaba Cloud
3

Enter credentials

  • API Key: Your API key
  • API Address: https://dashscope-intl.aliyuncs.com/compatible-mode/v1
4

Add provider

Click Add
5

Add models

Enter Model ID (e.g., qwen3.5-plus, qwen3.5-flash)

Embedding model setup (for knowledge bases)

To use local knowledge bases, add an embedding model:
1

Find Qwen Cloud

In Model Service section, find Qwen Cloud
2

Add embedding model

Add text-embedding-v4 as the embedding model
The multimodal embedding model multimodal-embedding-v1 is not supported in Cherry Studio.

Limitations

  • Thinking mode: Some Qwen3 models require thinking mode always on
  • Embedding: Only text-embedding-v4 supported, not multimodal embeddings

Examples

  • Basic chat
  • MCP tools integration
  • Local knowledge base
Enter a message to chat with Qwen models. Use the thinking mode toggle thinking mode for thinking-capable models like qwen3.5-plus or qwen3-max.
Cherry Studio chat interface

Troubleshooting

"enable_thinking is restricted to True"
Solution: Some Qwen3 models require thinking mode always on. Either update Cherry Studio or keep thinking mode enabled for these models.
Charges despite free quota
Solution:
  • Free quota is region and model specific
  • Each model has independent quota (qwen-max ≠ qwen-max-latest)
  • Quota display updates hourly, not real-time
  • Check Free quota details for coverage
  • Enable Free quota only to prevent unexpected charges
Model not responding
Solution: Verify API key is valid and has quota. Check Model list for supported model IDs.
Knowledge base embedding fails
Solution: Ensure you've added text-embedding-v4 in Model Service section and have API quota.

Advanced features

Thinking mode models

Models like qwen3.5-plus and qwen3-max support thinking mode for complex reasoning:
  • Toggle thinking mode with the button in input box
  • Some models require thinking mode always on
  • View thinking process in responses

MCP server options

ModelScope provides various MCP servers:
  • Fetch: Web scraping and content extraction
  • File operations: Local file management
  • API connectors: External service integration
  • No self-hosting required

Knowledge base tips

  • All data stored locally on your machine
  • Supports PDF, TXT, MD, and web URLs
  • Use for documentation, research papers, or internal docs
  • Combine with MCP tools for powerful workflows