Alibaba released its next-generation foundation model Qwen4-Max on Nov 16 and has made it available via API on the Qwen AI platform. Official disclosures and third-party evaluations show the model leading among Chinese models on several Arena leaderboards, but its overall ranking remains mid-to-low, prompting debate among users over its real competitiveness.

Segment leaderboards: No. 1 in China on Agent Arena

On Agent Arena, Qwen4-Max ranks No. 1 in China and No. 14 globally. The leaderboard mainly tests a model's agent capabilities — its ability to autonomously call tools and complete tasks step by step.

The model also ranks No. 6 on Code Arena WebDev and Image-to-WebDev Arena, and entered the top five on DesignArena. Those results are concentrated in specific task areas such as agents and code generation.

Leaderboard / metricQwen4-Max performance
Agent ArenaNo. 1 in China, No. 14 globally
Code Arena WebDevNo. 6
Image-to-WebDev ArenaNo. 6
DesignArenaTop five
Artificial Analysis Intelligence Index43 points, up 9 from the previous generation
Per-task cost$3.27, down 67%
Artificial Analysis overall leaderboard38/81

The Artificial Analysis Intelligence Index gives Qwen4-Max a score of 43, up 9 points from the previous generation. On cost, per-task cost fell 67% to $3.27, extending Alibaba's push on inference economics.

Mid-to-low overall, split user feedback

Despite the strong segment results, Qwen4-Max ranks only 38/81 on the Artificial Analysis overall leaderboard, a mid-to-low position that contrasts with its leading spots on some leaderboards.

One user on the V2EX forum asked for real-world test data, saying they "already know it's pretty bad" but still wanted to know whether it beats the previous Qwen3.8-Max, adding that "the unit price is cheaper after all."

Some users consider Qwen4-Max disappointing, with its overall ranking contrasting with its leading position on some leaderboards.

Feedback on actual Qwen4-Max experience remains scattered, with users focused on the balance between the size of the performance gain and the price advantage.

What to watch: real-world performance and adoption

The Qwen4-Max API is live, and its real-task performance awaits verification by more developers. The industry is watching whether the model can deliver on its leaderboard results in complex tasks, and whether its cost advantage translates into actual adoption.

With domestic foundation models iterating faster, market feedback on Qwen4-Max will be an important reference for gauging Alibaba's model competitiveness.