AI Disruption

AI Disruption

Qwen3-Max-Thinking: Alibaba’s Top Qwen 3 Reasoner

Meet Qwen3-Max-Thinking: Alibaba's 100% AIME scorer, now in early preview on Qwen Chat.

Meng Li's avatar
Meng Li
Nov 04, 2025
∙ Paid
7
4
Share

“AI Disruption” Publication 8100 Subscriptions 20% Discount Offer Link.


Alibaba’s “Super-Sized” Reasoning Model Finally Unveiled!

Just now, Alibaba released an early preview version of Qwen3-Max-Thinking, the most powerful reasoning model in the Qwen 3 series. Although this model is only a checkpoint from intermediate training, it has already achieved 100% accuracy on challenging reasoning benchmarks such as AIME 2025 and HMMT (Harvard-MIT Mathematics Tournament) through tool usage and test-time computation.

Image

However, this model actually received an “advance preview” on Lin Junyang’s (head of Tongyi Qianwen) personal social media account: at 11:54 PM on November 2nd, Lin posted “It’s here, you can try it,” accompanied by an image of Qwen3-Max with thinking mode enabled.

This post is for paid subscribers

Already a paid subscriber? Sign in
© 2025 Meng Li
Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture