Alibaba Cloud Model Studio:Claude Code

Supported models

Model Studio's Anthropic API-compatible service supports:

For model parameters and billing rules, see the Model list.

Recommendations

Recommended models by task type:

• Complex inference tasks (such as designing application architectures, implementing complex algorithms): Qwen-Plus series (upgraded to Qwen3.5), Qwen-Max series, qwen3-coder-next (combines coding capabilities with response speed), and qwen3-coder-plus.

• Simple coding tasks (such as writing functions, generating code comments): Qwen-Flash series and qwen3-coder-next.

1. Activate Model Studio

1. Create an account: If you do not have an Alibaba Cloud account, first create one.

   If you encounter issues, see Create an Alibaba Cloud account.

2. Activate Model Studio: Use your Alibaba Cloud account to go to the Model Studio console. After you read and agree to the terms, Model Studio automatically activates. If the Terms of Service do not appear, the service is already activated.

   If a message "You have not completed identity verification" appears, first complete identity verification.

First-time users can get a free quota (valid for 90 days). For details, see the Free quota for new users page.

2. Install Claude Code

npm install -g @anthropic-ai/claude-code

3. Configure base URL, API key, and model

Configure these environment variables to connect Claude Code to Model Studio:

1. Set ANTHROPIC_BASE_URL to https://dashscope-intl.aliyuncs.com/apps/anthropic.

2. Set ANTHROPIC_API_KEY or ANTHROPIC_AUTH_TOKEN to your Model Studio API key.

   ANTHROPIC_API_KEY or ANTHROPIC_AUTH_TOKEN can be used for access authentication. Set only one. The examples below use ANTHROPIC_API_KEY as an example.

3. Set ANTHROPIC_MODEL to a supported model from the Model list.

4. Run Claude Code

Navigate to your project directory and run claude:

claude

On first launch, Claude Code may prompt you to log in to an Anthropic account. Skip this step as follows:

1. Open the .claude.json file in your home directory:

   • macOS / Linux: ~/.claude.json

   • Windows: C:\Users\%USERNAME%\.claude.json

2. Set the value of the hasCompletedOnboarding field to true.

   {
     "hasCompletedOnboarding": true
   }

3. Save the file, then rerun claude in a new terminal.

(Optional) Additional model configuration methods

Claude Code supports several ways to configure models, listed from highest to lowest priority:

1. During a conversation: Use /model <model_name> to switch models. This is useful for temporary changes.

   /model qwen3.5-plus

2. When starting Claude Code: Use claude --model <model_name> to specify a model for a single session.

   claude --model qwen3.5-plus

3. Set environment variables: Assign different models to tasks of varying complexity. Claude Code automatically selects the appropriate model. This applies globally.

   export ANTHROPIC_DEFAULT_OPUS_MODEL="qwen3.5-plus"
   export ANTHROPIC_DEFAULT_SONNET_MODEL="qwen3.5-plus"
   export ANTHROPIC_DEFAULT_HAIKU_MODEL="qwen3-coder-next"

   where:

   • ANTHROPIC_DEFAULT_OPUS_MODEL: For difficult tasks, such as complex inference and architecture design.

   • ANTHROPIC_DEFAULT_SONNET_MODEL: For daily tasks, such as writing code and implementing features.

   • ANTHROPIC_DEFAULT_HAIKU_MODEL: For simple tasks, such as syntax checks and file searches.

4. Set permanently in settings.json: Create a settings.json file in the project root or user home directory with the model configuration. This creates a permanent setting.

   {
     "env": {
       "ANTHROPIC_DEFAULT_OPUS_MODEL": "qwen3.5-plus",
       "ANTHROPIC_DEFAULT_SONNET_MODEL": "qwen3.5-plus",
       "ANTHROPIC_DEFAULT_HAIKU_MODEL": "qwen3-coder-next"
     }
   }

Question 1: After starting Claude Code, the interface shows “Unable to connect to Anthropic services.” What should I do?

Reconfigure the environment variables, then run claude again.

Question 2: Tokens are consumed too quickly. How to save tokens?

Claude Code scans your project directory, reads code files, and maintains conversation history, consuming significantly more tokens than standard chat. These methods help reduce token usage:

1. Remove unrelated files: Start Claude Code from your specific project directory with only essential files. Scanning irrelevant files wastes tokens.

2. Summarize the conversation: Claude Code includes past messages as context. At 95% of the context window, it automatically summarizes the history. To summarize manually, type /compact.

3. Use precise instructions: Vague requests trigger unnecessary file scans. Ask clear, specific questions to minimize token waste.

4. Break tasks into smaller steps: For complex tasks, split them into several simple jobs.

5. Reset the context: Before starting a new task, type /clear to reset the context and avoid wasting tokens on irrelevant history.

For more token-saving tips, see the official Claude Code documentation. For token usage and expense details, see Bill query and cost management.