Key Takeaway
As of April 13, 2026, both HappyHorse-1.0 and xAI's Grok Imagine Video are claiming top positions on the Artificial Analysis Video Arena. The leaderboard is highly dynamic, and rankings may shift depending on the category and time of access.
The Situation
The AI video generation leaderboard has become a two-front battle. HappyHorse-1.0, developed by Alibaba's ATH division, topped the Artificial Analysis Video Arena in early April 2026. Shortly after, xAI's Grok Imagine Video also began claiming the #1 position.
According to WaveSpeed AI (April 12, 2026), "Grok Imagine Video recently ranked #1 on the Artificial Analysis Video Arena for both text-to-video and image-to-video generation, outperforming Runway Gen-4.5, Sora 2 Pro, and Google Veo 3.1."
Meanwhile, Atlas Cloud (April 13, 2026) reports HappyHorse-1.0 at 1384 Elo in text-to-video and 1413 Elo in image-to-video — with the latter being the highest score ever recorded on the platform.
What We Know About Each Model
| Feature | HappyHorse-1.0 | Grok Imagine Video |
|---|---|---|
| Developer | Alibaba ATH AI | xAI (Elon Musk) |
| Model Type | Open source | Closed / proprietary |
| Architecture | Unified Transformer (15B params) | Not publicly disclosed |
| Native Audio | Yes (one-pass generation) | Not confirmed |
| Resolution | Native 1080p | Not publicly specified |
| Reference-to-Video | Not confirmed | Yes (multi-image reference inputs) |
| API Status | Coming April 30 | Available through xAI / Grok platform |
| Self-hosting | Yes (open weights on GitHub) | No |
Why Rankings May Differ
The Artificial Analysis Video Arena uses an Elo-based ranking system derived from blind user comparisons. Rankings can fluctuate based on several factors:
- Category differences: Artificial Analysis operates four separate leaderboards — text-to-video (with and without audio) and image-to-video (with and without audio). A model can be #1 in one category but not others.
- Time sensitivity: Elo scores update continuously as new comparisons are submitted. Rankings from one hour may differ from the next.
- New model entries: When a new model enters the arena, it can rapidly shift the rankings as it accumulates comparison data.
Different Strengths
Based on available information, the two models appear to have different strengths:
HappyHorse-1.0 excels in character depiction and portrait generation. According to Atlas Cloud, over 60% of test samples where HappyHorse was preferred were portraits or talking-head clips. Its unified audio-video generation is also a distinctive capability.
Grok Imagine Video appears to differentiate through its reference-to-video capability, which allows users to provide multiple reference images to guide video generation — a feature most competitors limit to 4 or fewer inputs.
Bottom Line
The #1 position on the Artificial Analysis Video Arena is no longer a settled question. Both HappyHorse and Grok Imagine Video have strong claims, and the leaderboard is likely to continue shifting as both models accumulate more comparison data and as new competitors enter the arena.
For developers, the key differentiator may be access model: HappyHorse is open-source and self-hostable, while Grok Imagine Video is available through xAI's platform. The choice depends on whether you prioritize control and customization (HappyHorse) or immediate availability and reference-to-video features (Grok).
Sources
- Atlas Cloud — "HappyHorse-1.0 Takes First Place, API Coming Soon" (April 13, 2026) atlascloud.ai
- WaveSpeed AI — "Introducing xAI Grok Imagine Video" (April 12, 2026) wavespeed.ai
- Artificial Analysis — Video Arena Rankings (accessed April 13, 2026) artificialanalysis.ai