AI modelsNovember 16, 2026
-22
š§
Alibaba Strikes Back with Qwen4-Max
Alibaba has released Qwen4-Max, a new flagship model reported to be more capable and cheaper than its predecessor. It reportedly ships with around 2.4 trillion parameters and a revamped native-vision MoE architecture, though it still trails Moonshot's Kimi K3.1 Pro by a few points.
#Alibaba#Qwen#MoE#LLM#Benchmarks

š„ What happened
Alibaba released Qwen4-Max, its new flagship model, now generally available via API after a preview on November 11. It uses a revised Mixture-of-Experts architecture with 2.4 trillion parameters, but only 71 billion active per token ā 25% fewer than its predecessor.
š” Why it matters
Qwen4-Max scores 43 points on Artificial Analysis' Intelligence Index, nine more than Qwen3.8-Max, while costing just $3.27 per task ā a third of the previous model. That puts it just behind Chinese rival Kimi K3.1 Pro (46 points) but well ahead of Tencent Hy4 (35 points).
ā” Our take
Alibaba is optimizing for cost, not peak performance: Qwen4-Max is cheaper, but Western models like Claude Fable 5.2 (65 points) and Gemini 4 Carbon (60 points) remain far ahead. If you're budget-conscious, you win ā if you want the best, you pay elsewhere.
The title, summary and analysis of this item were produced automatically by an AI system and have not been editorially reviewed. They may contain errors, bias or omissions ā when in doubt, read the linked original source.
Deep Dives & Similar Intelligence
Alibaba open-sources its code review agent
Tools & ProjectsNovember 13, 2026

Tencent's Hy4 Flash Slashes AI Costs
AI modelsNovember 16, 2026
ki-daily.
OpenAI slashes prices, dominates benchmarks
AI modelsNovember 17, 2026

165M Model Trained for $4,200
ResearchNovember 15, 2026

Microsoft VP ships GPU task manager
Tools & ProjectsNovember 12, 2026