Breaking

Alibaba's Wan-2.7 I2V hits No. 5 on Video Arena at Elo 1,434

June 24, 2026 at 23:22 EDT

Media Generation
Foundation Models

Alibaba's video model Wan-2.7 I2V has reached No. 5 on the Image-to-Video category of the community-voted "Video Arena" leaderboard, posting an Elo score of 1,434. Alibaba's Wan team disclosed the result in late June 2026. Rankings are decided through head-to-head voting, in which real users compare two outputs for a given task and pick the one they prefer, and are updated live on the Video Generation Arena Leaderboard run by ArtificialAnalysis.

June 23, 2026 · Alibaba Tongyi Lab

Wan-2.7 cracks the Top 5 in Image-to-Video

Alibaba's image-to-video model scores 1,434 and ranks 5th on a crowdsourced leaderboard built from blind, head-to-head community voting — placing it ahead of every Google Veo-3.1 variant.

Rank on the I2V leaderboard

1,434 ±8

Elo score from blind voting

14,113

Votes cast for this model

1.35M

Total votes · 42 models

Top of the Image-to-Video leaderboard

1,474

#1 Dreamina
Seedance 2.0

—

#2 Gemini
Omni Flash

—

#3 Grok
Imagine

—

#4 HappyHorse
1.0

1,434

#5 Wan-2.7
I2V

Bars scaled to relative Elo score; values shown where reported.

What 2.7 adds: controllability

Up to 5 references
character & voice consistency

First & last frame
set start and end frames

9-grid input
3×3 references at once

Instruction editing
change weather / clothing by text

720p / 1080p · 2–15 sec clips · multimodal control across text, image, audio and video · modes: T2V / I2V / R2V.

Well regarded

Gains in visual quality, motion and consistency
Improved audio sync
Production-grade control & text-instruction editing

Criticisms

Focus on control, not a pure image-quality leap
API-only — local execution is difficult
Slow generation

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.