Alibaba Announces HappyHorse-1.0, Another Multimodal Model Coming Soon
On April 10, Alibaba officially announced on social platforms that HappyHorse is its self-developed multimodal model product, while clarifying that all related domain names circulating online are not official websites. Later, third-party AI evaluation platform Artificial Analysis announced that the API interface for the HappyHorse model is planned to be officially opened to the public on April 30.
Model R&D Background and Team Affiliation
According to sources close to Alibaba from Yicai “Neocortex”, HappyHorse-1.0 was developed by the former Future Life Laboratory team under Taobao Tmall Group. In Alibaba’s latest round of organizational restructuring, this team has been incorporated into the newly established Alibaba Token Hub (ATH) business group, under the “AI Innovation Division”.
The source also revealed that HappyHorse-1.0 is only the first product in the team’s self-developed multimodal model matrix, and Alibaba will soon launch another multimodal model with different positioning to form a product梯队 with HappyHorse-1.0.
Impressive Performance in Evaluation
Last week, HappyHorse-1.0 first appeared in the anonymous voting test on the third-party evaluation platform Artificial Analysis. In the blind test where users did not know the identity of the model, HappyHorse-1.0 received more user preference votes than ByteDance’s Seedance 2.0 and Kuaishou’s Kling 3.0 among video generation models.
Specific performance:
- Pure video generation (no audio): HappyHorse-1.0 shows significant advantages, with user votes far exceeding competitors
- Integrated audio and video generation: Performance is comparable to Seedance 2.0, placing it in the first tier of the industry
Before this, Alibaba’s video generation model R&D was mainly led by the Wanxiang team under Tongyi Laboratory, which was just upgraded to Tongyi Large Model Division on April 8. The launch of HappyHorse-1.0 marks that the “AI Innovation Division” has become another highly competitive multimodal model R&D team under Alibaba’s ATH business group.
Closed-source Strategy and Industry Competition
It is reported that HappyHorse-1.0 will not be open source for the time being, which is consistent with Alibaba’s recent overall model strategy of gradually turning to closed source. Since the end of March, products such as the full-modal large model Qwen3.5-Omni, the first model of the Qwen3.6 series Qwen3.6-Plus, and the video generation model Wan2.7-Video released by Alibaba have not been open source.
Previously, Alibaba only adopted a closed-source strategy for the strongest Max series of its models, which users could not deploy independently and had to pay for through API calls. This strategy has now been extended to other versions beyond the Max series, indicating Alibaba’s strategic adjustment in AI commercialization.
Improving multimodal capabilities is one of Alibaba’s current strategic priorities. During the 2026 Spring Festival, the video generation model Seedance 2.0 launched by ByteDance was highly recognized by the market, and was evaluated by foreign media as a Chinese AI model that “puts Hollywood in panic”. Sources close to Alibaba revealed that the launch of ByteDance’s product once surprised Alibaba’s internal team, because at the Volcano Engine annual conference in December last year, the Seed1.8 base model and Seedance 1.5 pro video generation model released by Byte did not perform prominently. The core update of Seedance 1.5 pro was the launch of the “integrated audio and video generation” function, which Alibaba had already implemented in Wan2.5 released in September last year.
Changes in Market Landscape
Compared with pure text chat models, multimodal generation consumes more computing resources and Tokens, which will directly affect the market share of cloud computing service providers in the MaaS (Model as a Service) market. According to data released by IDC, Volcano Engine had occupied 49.2% of China’s MaaS market share in the first half of 2025, while Alibaba Cloud’s market share was 27%.
The launch of the HappyHorse series of models will become an important product matrix for Alibaba to compete with competitors in the field of multimodal AI.
Experience HappyHorse AI Generation Capabilities
Now you can directly experience the powerful video generation capabilities of the HappyHorse model on our platform, no need to wait for the API to open.