Mysterious video generation AI model 'HappyHorse-1.0' achieves world's highest performance in anonymous testing.
A model named ' HappyHorse-1.0 ' has appeared in the video generation AI performance ranking of the AI information analysis site 'Artificial Analysis,' and has taken the top spot, surpassing high-performance models from Google and ByteDance.
Text to Video Leaderboard - Top AI Video Models
Image to Video Leaderboard - Top AI Video Models
https://artificialanalysis.ai/video/leaderboard/image-to-video
Artificial Analysis operates a video generation AI ranking site called ' Video Arena .' When users access Video Arena, they are shown two videos generated by two different models based on the same prompt. Users play both videos and vote for the one they feel is 'faithful to the prompt and high quality.' The system does not display information about 'which AI model was used to generate the video' during the video generation process, allowing for the creation of a 'ranking of video generation AI models that humans perceive as high quality' while concealing the names of the models.
I felt the video on the right was of higher quality this time, so I'm voting for the video on the right.
When you vote, model information will be displayed in the upper right corner of the screen. In this test, we can see that the videos of those who voted were generated with 'HappyHorse-1.0', while the videos of those who did not vote were generated with 'Veo 3.1 Lite'. In Video Arena, user votes accumulate in this way to determine the rankings.
As of the time of writing, HappyHorse-1.0 is ranked number one in the 'Generate silent videos from text' category, surpassing Google's Veo series and ByteDance's Seedance series.
The information page for HappyHorse-1.0 is marked 'coming soon,' and it is unclear which organization developed the model. Furthermore, since Artificial Analysis allows registration under code names, it is also unclear whether HappyHorse-1.0 is the official name or a code name.
HappyHorse-1.0 also reigns supreme in the category of 'generating videos with audio from text.'
It also ranks first in the category of 'Generating silent videos from images.'
In the category of 'Generating videos with sound from images,' Seedance 2.0 came in first place, with HappyHorse-1.0 coming in second, just one point behind.
The Artificial Analysis X account has published several examples of HappyHorse-1.0.
Example generations from HappyHorse-1.0 compared to Dreamina Seedance 2.0, Kling 3.0 Pro, grok-video-imagine and PixVerse V6 (Text to Video with Audio):
pic.twitter.com/GjtiY0COYl — Artificial Analysis (@ArtificialAnlys) April 7, 2026
Prompt [1/4]: A hula hoop spinning on a kid's waist, gradually climbing to their chest, then dropping to knees, then…
Although no information is publicly available on the Artificial Analysis page at the time of writing, there is information circulating on the internet that 'HappyHorse-1.0 was developed by Taotian Group, a subsidiary of Alibaba.'
Alright, now I can say it. From Alibaba's Taotian Group. (Alibaba has six major divisions, among which Qwen, Wan, and Zimage all come from the Tongyi Lab, which belongs to Alibaba Cloud, their relationship is somewhat like Amazon and AWS.) https://t.co/XvFloZpUUC pic.twitter.com/8mUXAq9LoQ
— Seiryu Saint (@bdsqlsz) April 8, 2026
Additionally, while several pages resembling the 'official HappyHorse-1.0 website' have been published online, many of them appear to be unrelated to HappyHorse-1.0. Therefore, it's best to refrain from registering as a member until information is released through Artificial Analysis or other means.
Related Posts:
