top of page

Alibaba’s new AI video model takes VLM top slot

Writer's picture: Patrick LawPatrick Law

The AI video generation space is evolving fast, and the latest disruptor is Alibaba’s Wan2.1—a model that just outperformed OpenAI’s Sora on VBench, the industry’s top AI video benchmark.


What is Wan2.1?

Alibaba’s Wan2.1 is an open-source AI video model that topped VBench by excelling in:

Motion Smoothness – More natural, fluid movements.

Text Rendering – Better accuracy with readable text.

Physics Accuracy – More realistic object interactions.


This game-changing model sets a new bar for AI video realism. You can explore more about it at WanX AI.


Wan2.1 vs. OpenAI’s Sora

Both are top-tier AI video generators, but here’s how they compare:

Feature

Alibaba’s Wan2.1

OpenAI’s Sora

Model Type

Open-source

Proprietary (Closed-source)

Benchmark Ranking (VBench)

Lower ranking

Motion Smoothness

Superior

Slightly less refined

Text Rendering

More accurate

Occasional distortions

Physics Accuracy

Leading in realism

Strong but less precise

Availability

Free for developers

Limited access

Why Open-Source Wins

Unlike Sora, Wan2.1 is open-source, giving developers access to its code and weights, allowing:

  • Faster innovation through collaboration.

  • Customization for specific needs.

  • More transparency in AI development.


What This Means for AI-Generated Video

With Alibaba leading the charge, we could see:

  • Higher quality AI-generated films, ads, and animations.

  • More accessible AI tools for creators and developers.

  • A shift from closed AI models to open innovation.


Which one do you think leads the future of AI video? Let us know! 🚀




 
 
 

Comments


bottom of page