A frontier mixture-of-experts model tuned for coding and office productivity.

Qwen3.8
Qwen3.8 Max
2.4T MoE flagship with native vision6% offAPI formatOpenAI
A 2.4-trillion-parameter MoE flagship with native vision understanding for coding, office, and professional tasks.
qwen3.8-maxModel details
Capabilities, benchmarks, and typical use
Qwen3.8 Max is Alibaba's 2.4-trillion-parameter MoE flagship model. It strengthens coding and office scenarios and targets professional work such as legal, financial, and design tasks. It has native vision understanding, parsing long documents and long videos, and supports autonomous planning and iteration for long-horizon tasks. The model accepts text, image, and video inputs and returns text. It exposes a 1M-token context window with a maximum input of 991,808 tokens and a maximum output of 131,072 tokens, and a maximum chain-of-thought length of 262,144 tokens in thinking mode. Function calling, structured output, web search, prefix continuation, and context caching are supported. The original prices follow the active local new-api configuration of $1.60 per 1M input tokens, $4.80 per 1M output tokens, and $0.20 per 1M cache-read tokens.
Reads images and long videos alongside text for charts, documents, and multimodal perception.
Plans and iterates independently across multi-step tasks with a full tool ecosystem.

