Alibaba made Qwen4-Max generally available on QwenCloud today, after a limited preview that started November 11.
Alibaba made Qwen4-Max generally available on QwenCloud today, after a limited preview that started November 11. It's a mixture-of-experts model with a 2.4 trillion-parameter backbone, 71 billion parameters active per token, native vision, and a context window of up to 1 million tokens.
The pricing is the part I keep coming back to. At international list rates it costs $1.50 per million input tokens and $4.50 per million output, with cached input at $0.075. FourWeekMBA did the math from both public pricing pages and puts that output price 91% below OpenAI's listed $50 per million for GPT-6 Astra. It's also about 25% cheaper than Alibaba's own prior Qwen3.8-Max, which was $2/$6.
On Alibaba's Work360 benchmark, average task cost fell from $4.43 to $1.67, roughly a 60% drop. That's their benchmark, so I'd want independent numbers before leaning on it too hard. Still, the direction is clear: Alibaba shipped a newer flagship and charged less for it.
The Max model stays closed-weight, though Alibaba says smaller Qwen4 models are headed for open source. For anyone running agent workloads that chew through output tokens, I think the question gets uncomfortable quickly. How long can a team justify paying roughly ten times more per output token, and what would the US model have to deliver to earn that gap? $BABA
Sources:
TechNode: https://technode.com/2026/11/16/alibaba-launches-qwen4-max-with-lower-api-prices-and-a-million-token-context-window/
el-fondo.com: https://www.el-fondo.com/en/news/alibaba-launches-qwen4-max-ai-model-with-24t-parameters
FourWeekMBA: https://fourweekmba.com/ai-qwen4-max-launches-at-4-50-m-output-91-below-gpt-6-astra/
Not financial advice. For informational and educational purposes only.