Can DeepSeek withstand the V4 shockwave? Ask DaiDamao (代达劢)

Alibaba’s surprise contender

Alibaba (阿里巴巴) ATH has confirmed that HappyHorse is a model developed by its ATH innovation business unit and is currently in internal testing, with an API to be opened soon, it has been reported. The disclosure follows a flurry of attention after an anonymous entry labeled HappyHorse‑1.0 reportedly shot to the top of the Artificial Analysis AI Video Arena leaderboard, overtaking ByteDance (字节跳动)’s Seedance 2.0, Kuaishou (快手)’s 可灵AI and Google’s Veo 3 Fast. Short and sharp: a previously unnamed video model has forced big players back into the ring.

What it means for China’s video‑AI race

For Western readers: ATH is Alibaba’s research and innovation arm pushing into generative and video AI, a sector increasingly crowded by domestic platforms that can no longer be treated as second‑tier to Silicon Valley. The leaderboard surge, if sustained, would signal that China’s homegrown model architectures and dataset scale are competitive even against global incumbents. But rankings are one thing, deployment and safeguards another. Who controls the API, and how models are fine‑tuned for content moderation and local norms, will matter as much as raw Elo scores.

Geopolitics, uncertainty and next moves

It has been reported that the ranking data come from Artificial Analysis’ Video Arena, but evaluation setups vary and benchmarking can be fluid. Geopolitical factors — export controls on advanced chips and shifting trade policy — remain background noise that shapes how quickly companies can scale model training and inference. Online commentators, reportedly including handles such as “DaiDamao (代达劢),” are already asking whether rivals like DeepSeek can absorb this V4‑era shockwave and respond at scale. For now, HappyHorse is a headline; proving durability will take live traffic, transparent evaluation and time.