Qwen3.8 Max

Qwen3.8-Max is Alibaba's flagship Qwen mixture-of-experts model. It accepts text, images, and video and returns text; the provider reports 2.4 trillion total parameters with 95 billion active.

It is especially suited to advanced coding, long-horizon agent workflows, complex professional work, and multimodal analysis of long documents or videos.

Its strengths include a 1M-token context window, optional thinking, function calling, structured outputs, built-in tools such as web search and code interpretation, and native visual understanding.

Flagship capability can cost more and take longer than lighter Qwen tiers. Long agent runs may compound errors and spending, so set budgets, add checkpoints, and review high-stakes outputs.

Choose qwen3.8-max in TokenHub's model field, then send requests to your TokenHub endpoint with your account credentials. Confirm the supported parameters, input formats, and limits shown in your TokenHub account before production use.

As of August 2026, QwenCloud lists qwen3.8-max for API use at $2 per 1M input tokens, $6 per 1M output tokens, and $0.25 per 1M implicit-cache input tokens. TokenHub availability and billing may differ, so check its live model page.

Choose it when you need Qwen's highest capability for complex coding, multimodal reasoning, or long agentic work. For cost-sensitive workloads, compare qwen3.7-plus or qwen3.7-flash and evaluate quality, latency, and cost on representative tasks.