Alibaba claims HappyHorse-1.0 and will also release another multimodal model

According to a report from CryptoWorld News, on April 10, Alibaba officially confirmed that the video generation model HappyHorse-1.0 is a self-developed product. The model originated from the former Future Life Laboratory team of Taotian Group. In Alibaba’s latest organizational restructuring, this team was placed under the newly established Alibaba Token Hub (ATH) business group, in its “AI Innovation Division.” In an anonymous vote on the third-party evaluation platform Artificial Analysis, HappyHorse-1.0 received significantly more votes than ByteDance’s Seedance 2.0 and Kuaishou’s Kling 3.0 in pure video generation tasks; in integrated audio-visual generation, it performed comparably to Seedance 2.0. Sources close to Alibaba revealed that HappyHorse-1.0 is only one of several multimodal models developed by this team. In the near term, Alibaba will also roll out another different multimodal model. At present, HappyHorse-1.0 is not open source, which is consistent with Alibaba’s recent overall shift to closed-source model strategies—since the end of March, multiple new models released by Alibaba have not been open sourced. The rapid push for multimodal models is partly driven by the fact that ByteDance’s Seedance 2.0’s outstanding performance during the 2026 Spring Festival had surprised Alibaba internally, and that multimodal generation will significantly increase token consumption, thereby affecting the market share of MaaS (Model as a Service). According to IDC data, in the first half of 2025, Volcano Engine had already captured 49.2% of that market, while Alibaba Cloud accounted for only 27%.