Qwen4-Max

Qwen4-Max is Alibaba's top-tier proprietary model, released July 11, 2026, sitting above the open-weight Qwen3 family the company publishes on Hugging Face. Unlike those, Qwen4-Max is closed and reachable only through Alibaba Cloud's API, Qwen Chat, or partner platforms. It accepts text and image input, carries a 1-million-token context window, and is priced at $1.20 per million input tokens and $4.80 per million output tokens. Alibaba positions it directly against GPT-5.6 and Gemini 3.5 Pro in its own benchmark comparisons, and it holds an LMArena score around 1352.

The honest picture: the split between open Qwen3 weights and closed Qwen4-Max is a deliberate strategy — capable mid-size weights build developer mindshare while the largest model stays behind a paid API. Parameter count is undisclosed, so capacity claims rest on vendor benchmarks alone. Teams evaluating it should benchmark against their actual workloads rather than the marketing comparisons.