Alibaba (Qwen)

Alibaba's Qwen family offers strong multilingual models with excellent coding and reasoning capabilities. Qwen3.8 Max is a 2.4 trillion parameter model optimized for full-stack development, data analysis, and complex workflows.

Setup

# 1. Get your API key from https://dashscope.console.aliyun.com/
export DASHSCOPE_API_KEY=sk-...

# 2. Verify detection
arcana doctor

# 3. Run
arcana run "hello"

Available Models

Highlights

• 2.4T parameters — Qwen3.8 Max is one of the largest open-weight models
• Strong coding — competitive with GPT-4 and Claude on code generation tasks
• Multilingual — excellent support for Chinese, English, and 29+ other languages
• Long context — up to 1M tokens for analyzing large codebases and documents
• Cost-effective — significantly cheaper than Western frontier models

Config File

{
  "provider": "alibaba",
  "model": "qwen3.8-max"
}

Environment Variables

Via OpenRouter

You can also access Qwen models through OpenRouter with a single key:

export OPENROUTER_API_KEY=sk-or-...
# Then use qwen3.8-max directly — Arcana routes through OpenRouter

Pricing Notes

Qwen models are very affordable. Approximate costs per 1M tokens via DashScope:

Tips

• qwen3.8-max is excellent for full-stack development at a fraction of GPT-4 cost
• Qwen models have strong multilingual capabilities — great for international projects
• Use as a utility model for extraction, summarization, and data processing tasks
• Qwen models are OpenAI-compatible — they work with OpenAI client libraries
• Great choice for batch processing and high-volume workloads due to low pricing