Connect OpenClaw to use models - Alibaba Cloud Model Studio

Supported models

OpenClaw connects to Model Studio through an OpenAI-compatible interface and supports the following models:

Category	Models
Qwen (proprietary)	Qwen Max, Qwen Plus, Qwen Flash, Qwen Coder
Qwen (open source)	Qwen3, QwQ-Open Source, QwQ-Preview, Qwen2.5, Qwen-Coder
Third-party	DeepSeek, Kimi, GLM

Third-party text generation models are available only in China (Beijing). Use that region's base URL and API key. See Step 3: Configure a Model Studio model.

For frequent use, the Coding Plan offers a monthly subscription with up to 90,000 requests.

Prerequisites

Before you begin, make sure you have:

An Alibaba Cloud account with identity verification completed
Node.js 22 or later installed (download)
Model Studio activated (see Step 1)

Configuration steps

Step 1: Activate Model Studio

Log on to the Model Studio console with your Alibaba Cloud account.
Read and agree to the Terms of Service to activate Model Studio. If they don't appear, you've already activated it.
Claim your free quota for new users (valid for 90 days).

Important

After the free quota expires or runs out, standard billing applies. Turn on Free quota only to prevent unexpected charges. Check the console for current pricing.

Step 2: Install OpenClaw

Check your Node.js version:

node --version

If you have a version below 22, upgrade Node.js before proceeding.

Installation methods

macOS and Linux

Use the installation script:

curl -fsSL https://openclaw.ai/install.sh | bash

Or install with npm:

npm install -g openclaw@latest

Windows

Run in PowerShell:

iwr -useb https://openclaw.ai/install.ps1 | iex

Or install with npm:

npm install -g openclaw@latest

Complete the setup wizard

OpenClaw starts a configuration wizard after installation. To run it manually later, use openclaw onboard.

Select the following to skip provider configuration for now and to set it up in the next step:

Prompt	Select
I understand this is powerful and inherently risky. Continue?	Yes
Onboarding mode	QuickStart
Model/auth provider	Skip for now
Filter models by provider	All providers
Default model	Keep current
Select channel (QuickStart)	Skip for now
Configure skills now? (recommended)	No
Enable hooks?	Press Space to select, then press Enter
How do you want to hatch your bot?	Do this later

Step 3: Configure a Model Studio model

Warning

Set "reasoning": false for every model. If reasoning is true or omitted, all responses will be empty.

Connecting OpenClaw to Model Studio requires three parameters from the same region:

Parameter	Description
Base URL	The API endpoint for your region (see table below)
API key	Model Studio access credential for authentication and billing
Model ID	The model to use, such as `qwen3.5-plus` or `qwen3-coder-next`

Base URLs and API key management by region

Region	Base URL	API key management
China (Beijing)	`https://dashscope.aliyuncs.com/compatible-mode/v1`	Key Management (Beijing)
Singapore	`https://dashscope-intl.aliyuncs.com/compatible-mode/v1`	Key Management (Singapore)
US (Virginia)	`https://dashscope-us.aliyuncs.com/compatible-mode/v1`	Key Management (Virginia)

Supported models vary by region. For example, US (Virginia) doesn't currently support qwen3-coder-next, and third-party text generation models are available only in China (Beijing). For the full list, see Model list.

Method 1: Web console (recommended)

Start the web console:
```
openclaw dashboard
```
The browser opens http://127.0.0.1:18789. If it doesn't open automatically, navigate to this address.
In the navigation pane, choose Config > RAW.

Replace the "agents": {...}, section with the configuration below. Replace DASHSCOPE_API_KEY with your actual API key.

Important

This configuration is for pay-as-you-go billing. If you use the Coding Plan, see Coding Plan for OpenClaw for plan-specific settings.

This example uses qwen3.5-plus and qwen3-coder-next in Singapore. The base URL, API key, and models must belong to the same region.

To add other models, define them in providers.bailian.models (set reasoning to false) and add a "bailian/<Model ID>": {} entry to agents.defaults.models. See Model list for supported models per region.

  "models": {
    "mode": "merge",
    "providers": {
      "bailian": {
        "baseUrl": "https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
        "apiKey": "DASHSCOPE_API_KEY",
        "api": "openai-completions",
        "models": [
          {
            "id": "qwen3.5-plus",
            "name": "qwen3.5-plus",
            "reasoning": false,
            "input": ["text", "image"],
            "contextWindow": 1000000,
            "maxTokens": 65536
          },
          {
            "id": "qwen3-coder-next",
            "name": "qwen3-coder-next",
            "reasoning": false,
            "input": ["text"],
            "contextWindow": 262144,
            "maxTokens": 65536
          }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "bailian/qwen3.5-plus"
      },
      "models": {
        "bailian/qwen3.5-plus": {},
        "bailian/qwen3-coder-next": {}
      }
    }
  },

Click Save, then click Update to apply.

Method 2: Edit the config file

OpenClaw stores its configuration in ~/.openclaw/openclaw.json, which it reads at startup.

Open the config file in a text editor:
```
nano ~/.openclaw/openclaw.json
```

Paste the configuration below. Replace DASHSCOPE_API_KEY with your actual API key.

Important

This configuration is for pay-as-you-go billing. If you use the Coding Plan, see Coding Plan for OpenClaw for plan-specific settings.

This example uses qwen3.5-plus and qwen3-coder-next in Singapore. The base URL, API key, and models must belong to the same region.

To add other models, define them in providers.bailian.models (set reasoning to false) and add a "bailian/<Model ID>": {} entry to agents.defaults.models. See Model list for supported models per region.

{
  "meta": {
    "lastTouchedVersion": "2026.2.1",
    "lastTouchedAt": "2026-02-03T08:20:00.000Z"
  },
  "models": {
    "mode": "merge",
    "providers": {
      "bailian": {
        "baseUrl": "https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
        "apiKey": "DASHSCOPE_API_KEY",
        "api": "openai-completions",
        "models": [
          {
            "id": "qwen3.5-plus",
            "name": "qwen3.5-plus",
            "reasoning": false,
            "input": ["text", "image"],
            "contextWindow": 1000000,
            "maxTokens": 65536
          },
          {
            "id": "qwen3-coder-next",
            "name": "qwen3-coder-next",
            "reasoning": false,
            "input": ["text"],
            "contextWindow": 262144,
            "maxTokens": 65536
          }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "bailian/qwen3.5-plus"
      },
      "models": {
        "bailian/qwen3.5-plus": {},
        "bailian/qwen3-coder-next": {}
      }
    }
  },
  "gateway": {
    "mode": "local",
    "auth": {
      "mode": "token",
      "token": "test123"
    }
  }
}

Save the file:
- nano: Press Ctrl+X, then Y, then Enter.
- vim: Press Esc, type :wq, then press Enter.
Verify the saved configuration:
```
cat ~/.openclaw/openclaw.json
```

Step 4: Start a conversation

Terminal (TUI)

openclaw tui

OpenClaw TUI

Web interface

openclaw dashboard

OpenClaw web interface

Switch models

In the current session (temporary)

Open the TUI with openclaw tui, then run:

/model qwen3-coder-next

The interface displays "model set to qwen3-coder-next" as confirmation. This change applies only to the current session.

As the default model (permanent)

Update the agents.defaults.model.primary field in your config file:

{
    "agents": {
        "defaults": {
            "model": {
                "primary": "bailian/qwen3.5-plus"
            }
        }
    }
}

Replace qwen3.5-plus with your desired model ID. This applies to all future sessions.

FAQ

Model not found or empty response

Check three things:

Model ID spelling. The model ID must match exactly. For example, qwen3.5-plus, not qwen-3.5-plus.
Provider name consistency. If the provider is named bailian in your configuration, reference the model as bailian/qwen3.5-plus -- not just qwen3.5-plus.
Thinking mode disabled. The reasoning parameter must be false for every model. If reasoning is true or omitted, responses will be empty.

"command not found" when running an OpenClaw command

First, verify the installation:

openclaw --version

If a version number appears, installation succeeded. The issue is likely an unmigrated command from the rename. Try clawdbot or moltbot instead of openclaw for that command. Alternatively, reinstall the latest version:

npm install -g openclaw@latest