Alibaba
Latency—
Throughput—
Upload rate—
Context length991K tokens
Max output128K tokens
Input price$2per 1M tokens
Output price$6per 1M tokens
Cache read$0.25per 1M tokens
Cache write$2.5per 1M tokens
/alibaba/qwen3.8-max
Qwen3.8-Max is a flagship multimodal reasoning model from Alibaba. Built on a sparse MoE architecture with 2.4T total and 95B active parameters, it is designed for long-horizon autonomous work — multi-day software engineering runs, end-to-end research reproduction on GPUs, and closed-loop execution where plans are revised against real feedback rather than answered once. It is suited for agentic coding and self-evolving harnesses, deep research and quantitative analysis, GUI and desktop operation, hundred-page document and full-codebase comprehension, visual reasoning from screenshots to working UI reconstruction, long-video understanding, and complex real-world professional deliverables. With a 1M context window, three thinking levels, and OpenAI/Anthropic-compatible interfaces, it drops directly into Claude Code, Codex, and Qoder workflows.
curl ${API_BASE_URL}/v1/chat/completions \
-H "Authorization: Bearer <YOUR_API_KEY>" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen3.8-max",
"messages": [
{
"role": "user",
"content": "Hello"
}
]
}'Replace <YOUR_API_KEY> with the API key generated on your token management page.
All requests must include the Authorization: Bearer <TOKEN> header.
RPM = requests per minute, TPM = tokens per minute, RPD = requests per day. Limits apply by token group.