Live updates
Alibaba unveils Qwen4-Max, cutting token use by up to 50% and average cost per task by 62%
First seen on X 12 hours ago@Alibaba_Qwen ♥ 3,70035 views
Alibaba released Qwen4-Max, its new flagship model. The company says it beats Qwen3.8-Max while using up to about 50% fewer tokens, and that the average cost per task in its internal work evaluation fell 62%, from $4.43 to $1.67.
According to the Qwen blog and the official Qwen account on X on Nov 16, Qwen4-Max is a mixture-of-experts model on the new Qwen4 architecture, with 2.4 trillion backbone parameters and 71 billion active per token. It has native vision and a 1-million-token context window. API pricing is $1.50 per million input tokens, $4.50 per million output tokens and $0.075 per million for implicit caching, all below Qwen3.8-Max, with caching down 70%. Alibaba showed it building 3D scenes in Blender, games in Unreal Engine and car parts for machining, and said programming a single workpiece dropped from about 475 minutes to 24. Its score on Alibaba's in-house Work360 benchmark rose from 47.74 to 52.69. In Alibaba's own comparison chart, it scored 21.3 on the HLAB legal-work test, ahead of Anthropic's Opus 5.5 (9.2) and OpenAI's GPT-6 Astra (7.1), but trailed on Terminal-Bench 4.0 with 48.5 against Opus 5.5's 59.6 and Kimi K3.1's 67.2.
The API is live on QwenCloud, and the model is also available in Qwen Studio, QwenWork and Qoder. The Max weights are not open, but Alibaba said "the open weights of smaller Qwen4 models will be released." Arena said Qwen4-Max debuted at No. 6 on its Code Arena WebDev leaderboard with a preliminary 1,750 points, the top Chinese model there. The announcement passed 200,000 views in under three hours, and replies kept asking when a cheaper Qwen4 Flash would arrive.
“📢 Meet Qwen4-Max, your new AI teammate for work, creativity, and discovery!”
