Skip to content
AIMODELRACE
Alibaba QwenVideo modelOpen weights

Wan 2.7

Open-weights video with photoreal faces and native audio.

Video Generation

#4

of 18
6.1 pts behind #1

0

races led

0

weeks at #1

85.5

score

Released
Jun 20, 2026
4 mo ago
Context
—
n/a
Price
$0.05 / sec
list price
Speed
—
median
Modalities
text · image
outputs video, audio
Weights
Open
self-hostable

The 30-second version

Alibaba's open-weights video model with photoreal faces and native audio.

Key takeaways
  • 1Best open-weights video realism.
  • 2Native audio.
  • 3Self-hostable.
  • 4Needs serious GPUs.

What you need to know

Purpose

Open, customisable video generation.

Unique strengths

  • Open weights.
  • Faces.

Where it ranks

Current snapshot · illustrative scores · click any race to watch it

CinematicAnimationPhysicsAudioConsistency

Wan 2.7Best on each axis

Video Generation over time

#1 frontierWan 2.7Gen-4.5LTX-2.3
65.070.075.080.085.090.095.02025Apr ’25Jul ’25Oct ’252026Apr ’26Jul ’26Oct ’26

Best use cases

01

Private pipelines

Video generation on your own infrastructure.

When to choose it, and when not to

Update history

The distilled version first; open any release for the deeper breakdown.