HappyHorse-1.0 tops the AI video leaderboard, giving Alibaba's model stack a new catalyst
AI summary card
HappyHorse-1.0 tops the AI video leaderboard, giving Alibaba's model stack a new catalyst
Citi believes Alibaba's open-source video model HappyHorse-1.0, developed by Alibaba ATH, is leading the Artificial Analysis leaderboard, strengthening the case for Alibaba AI and cloud monetization while also intensifying competitive pressure on Kuaishou's Kling.
- HappyHorse-1.0 achieved ELO scores of 1,388 and 1,415 on the Artificial Analysis text-to-video and image-to-video leaderboards, respectively, ahead of mainstream models such as Seedance 2.0, Kling 3.0, and the Google Veo series.
- The report says the model was developed by a team under Alibaba ATH, and the creator name on the leaderboard has been changed from HappyHorse to Alibaba-ATH, supporting the view that it is linked to Alibaba's ecosystem.
- Citi believes video models require significantly more tokens; if HappyHorse gains popularity, it could translate into higher token monetization and cloud revenue for Alibaba.
- For Kuaishou, the emergence of HappyHorse means stronger competition for Kling in the AI video model race, although Kling still retains advantages in cinematic quality and its user base.
Report interpretation
Overview
This report focuses on China internet and AI video generation. The core event is the open-source video model HappyHorse-1.0 topping the Artificial Analysis AI Video Arena leaderboard around April 8, 2026. Citi views this as an important advance in Alibaba's model stack because China's video generation models had previously been dominated mainly by ByteDance and Kuaishou. The report also assesses the event's impact on Alibaba cloud revenue, token monetization, Kuaishou Kling's competitive landscape, and valuations of related internet companies.
Core views
The key view is that HappyHorse-1.0's leaderboard lead improves the market's perception of Alibaba's AI video generation capabilities and could create incremental commercialization opportunities for Alibaba Cloud; however, its future competitiveness still depends on pricing, compute requirements, update cadence, and the monetization path of the open-source project. For Kuaishou, Kling still has advantages in cinematic quality and its user base, but competition from Alibaba's open-source model is rising.
Analysis framework
The report mainly uses an event-driven and peer-comparison framework: it first cites information from 36Kr, Wall Street CN, and the Artificial Analysis leaderboard to confirm HappyHorse-1.0's ranking, technical characteristics, and linkage to Alibaba, then compares the user-perceived ELO scores of mainstream models such as Seedance, Kling, and Veo, and finally maps AI model progress to Alibaba Cloud revenue opportunities, competitive pressure on Kuaishou Kling, and SOTP valuations for related stocks.
Methodology notes
sum-of-the-parts valuation
Alibaba's target price is based on sum-of-the-parts valuation of the ecommerce, cloud intelligence, AIDC, and other businesses, with a 30% discount applied to net cash; Kuaishou's target price is composed of the value of its core ecommerce business and the valuation corresponding to Kling revenue.
user blind-test perception ranking
The Artificial Analysis leaderboard is not a pure technical-parameter score, but an ELO score aggregated from real-user blind tests, so it is more focused on perceived results after ordinary users view the outputs.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Alibaba Group Holding (9988.HK; BABA.N)primary beneficiary
- Strengths
- HappyHorse-1.0 is viewed as an important achievement in Alibaba's model stack; the high token consumption of video models may boost cloud revenue and monetization opportunities; segment valuation across ecommerce, cloud intelligence, AIDC and more provides SOTP support.
- Weaknesses
- The details of HappyHorse's capabilities, pricing, compute requirements, update frequency, and the monetization path of the open-source project remain unclear.
- Comparison
- In AI video generation, HappyHorse ranks ahead of parts of the leaderboard performance of Seedance 2.0, Kling 3.0, and the Google Veo series.
- Risks
- Failure to execute the new retail strategy, investment spending and margin pressure higher than expected, slowing user traffic and online GMV, M&A integration risk, macro and trade disputes, regulatory risk.
- Kuaishou (1024.HK)rising competitive pressure but still valuation support
- Strengths
- Kling has cinematic quality, a user base, and commercialization momentum; Kuaishou's ecommerce GMV is growing quickly.
- Weaknesses
- HappyHorse's success means Kling faces stronger competition in video models, which may weigh on its long-term revenue and valuation upside.
- Comparison
- The report places Kling 3.0 alongside HappyHorse-1.0, Seedance 2.0, and the Google Veo series as mainstream AI video generation models.
- Risks
- Competition from Douyin and video accounts platforms, slower ecommerce GMV growth, less-than-expected margin improvement, regulatory risk, and share sales by major shareholders.
Key data
- HappyHorse-1.0 text-to-video ELO1,388As of 2026-04-10, ranked near the top of the Artificial Analysis text-to-video leaderboard.
- HappyHorse-1.0 image-to-video ELO1,415As of 2026-04-10, ranked near the top of the Artificial Analysis image-to-video leaderboard.
- Seedance 2.0 text-to-video/image-to-video ELO1,273 / 1,359The report lists it as one of the mainstream competitors surpassed by HappyHorse-1.0.
- HappyHorse-1.0 model parameters15bnThe report says it uses a unified single-stream transformer architecture that supports joint video and audio output from mixed text/image inputs.
- Alibaba 9988.HK target priceHK$204Corresponding to the 2026-04-09 closing price of HK$122.9, implying about 66.0% upside.
- Alibaba BABA.N target priceUS$205Corresponding to the 2026-04-09 closing price of US$127.68, implying about 60.6% upside.
- Kuaishou 1024.HK target priceHK$72Corresponding to the 2026-04-09 closing price of HK$45.02, implying about 59.9% upside.
Impact & implications
If HappyHorse-1.0's performance and usage continue to improve, Alibaba could move from a demonstration of model capabilities to token consumption and cloud revenue conversion in AI video generation, strengthening market confidence in its AI infrastructure and application ecosystem. At the same time, the valuation premium and revenue expectations for Kuaishou Kling will need to face higher competitive intensity, and the industry landscape may evolve from a few leading models to a broader competition among more large-cap companies and open-source models.
Risks
- HappyHorse-1.0's leaderboard performance may not translate into sustained user usage, commercial revenue, or incremental cloud revenue.
- Pricing strategy, compute costs, and update cadence for the open-source model remain uncertain, which could affect Alibaba's subsequent monetization.
- Intensifying competition in the AI video generation industry could compress the growth space and valuation premium for existing models such as Kling.
- Alibaba faces downside risks from new retail execution, investment spending, margins, GMV, macro conditions, trade disputes, and regulation.
- Kuaishou faces risks from short-video and ecommerce competition, GMV growth, margin improvement, regulation, and share sales by shareholders.
What to watch
- Whether HappyHorse-1.0 continues to rank near the top of the Artificial Analysis leaderboard.
- Whether Alibaba ATH discloses more on model capabilities, pricing, compute requirements, and commercialization paths.
- HappyHorse's open-source ecosystem adoption rate, developer activity, and API/token usage.
- Whether Alibaba Cloud sees incremental contribution related to video generation model usage.
- Kuaishou Kling's product iteration, user growth, paid conversion, and performance gap versus HappyHorse.