Alibaba's AI Video Model Leaps to Global No. 2, Overtaking OpenAI Sora and ByteDance Seedance

Alibaba Cloud has launched HappyHorse 1.1, an enterprise-grade video generation model. Rather than a simple demo, this model is an API-first product that can be immediately integrated into corporate workflows, and to celebrate the launch, a 40% discount is being offered for two weeks.

The AI video market has recently undergone rapid change, with OpenAI's Sora being discontinued due to profitability issues and ByteDance's Seedance 2.0 indefinitely delaying its international launch due to copyright concerns. Amid this vacuum, HappyHorse has seized the opportunity to expand its market share on the strength of its technology.

According to benchmark results from the Artificial Analysis Video Arena, HappyHorse ranked at the very top in both text-to-video and image-to-video categories, surpassing Google's Veo-3.1 and xAI's Grok-Imagine-Video. This is not a mere statistical coincidence but a quality gap proven through actual user evaluations (Elo rating).

The technical core lies in a unified self-attention Transformer architecture with 15 billion parameters. By building an integrated system that processes text, image, video, and audio tokens as a single sequence, it handles all modalities in a single generation pass without the need for separate audio editing tools.

Key features of this 1.1 update are as follows:

• R2V (Reference-to-Video): A feature that uses multiple character reference images to consistently maintain the identity of a person within the video.
• Enhanced Motion Modeling: Improves upon the motion quality issues that were a limitation of previous models.
• Commercial Optimization: Practical optimization for advertising and content production, rather than for social media demos.