Skip to content

Qwen Provider

The default and recommended provider for Dial Code.

Quick Start

dial

On first run, your browser opens for Qwen OAuth authentication.

Free Tier

Qwen offers a generous free tier:

Limit Value
Requests per day 2,000
Requests per minute 60
Token limit None (per-request)
Cost Free

Authentication

OAuth Flow

  1. Run dial
  2. Browser opens to qwen.ai
  3. Log in or create account
  4. Authorize Dial Code
  5. Credentials cached locally

Re-authenticate

/auth

Clear Credentials

rm -rf ~/.dial/credentials

Available Models

Model Description
qwen3-coder Code-focused model
qwen3-coder-plus Enhanced code model
qwen-vl-max-latest Vision-capable model

Set default model:

{
  "model": "qwen3-coder-plus"
}

Vision Support

Qwen supports vision models for image analysis:

> [attach image] What's in this screenshot?

Configure auto-switching:

{
  "experimental": {
    "vlmSwitchMode": "once"
  }
}

Options:

Mode Behavior
once Switch for one query, then revert
session Switch for entire session
persist Stay with vision model

Regional Endpoints

International (Default)

Automatic - no configuration needed.

China (Mainland)

Use Alibaba Cloud Bailian:

export OPENAI_API_KEY="your_key"
export OPENAI_BASE_URL="https://dashscope.aliyuncs.com/compatible-mode/v1"
export OPENAI_MODEL="qwen3-coder-plus"

Or ModelScope (free tier):

export OPENAI_API_KEY="your_key"
export OPENAI_BASE_URL="https://api-inference.modelscope.cn/v1"
export OPENAI_MODEL="Qwen/Qwen3-Coder-480B-A35B-Instruct"

Troubleshooting

"Rate Limited"

  • Free tier: 60 requests/minute
  • Wait briefly and retry
  • Heavy use may trigger limits

"Authentication Failed"

# Clear and re-authenticate
rm -rf ~/.dial/credentials
dial

"Model Fallback"

During high load, the service may use fallback models. This is automatic and maintains service quality.

Next Steps