Sources: Alibaba anonymously released AI video model HappyHorse-1.0, which ranks at the top of Artificial Analysis' AI model leaderboard, above Seedance 2.0
Alibaba Group has anonymously released a new AI video generation model called HappyHorse-1., which rose to the top …
The InformationJuro Osawa
Context & Ripple Effects
Alibaba had already made image-and-video generation part of its open-model strategy through the global release of Wan 2.1, alongside broad Qwen-family releases. The reported appearance of a leading video model extends that strategy from publishing model families to testing frontier performance in a public comparison channel.
The anonymous approach also follows a nearby pattern: Zhipu reportedly used a temporary identity for an unrevealed model release on OpenRouter. Within Alibaba’s own product arc, the subsequent Token Hub effort around interactive 3D and video world generation suggests video capability is becoming a component of broader content and environment creation, not a standalone feature.
First-order effects
Alibaba gains a reported benchmark signal in AI video generation without initially attaching its brand, while Seedance 2.0 loses its leading position on the cited leaderboard.
Developers and enterprise evaluators have a new high-ranking video model to test, but the anonymous release makes provenance and support commitments less immediately clear than in a conventional product launch.
Second-order effects
Seedance and other video-model providers face pressure to demonstrate comparable quality in independent evaluations rather than relying only on launch claims.
Alibaba can use strong video-model performance to reinforce demand for its wider model ecosystem; its earlier open video release provides a distribution path if it chooses to formally publish or host the model.
Third-order effects
If anonymous benchmark releases become more common, public leaderboards may increasingly function as pre-launch testing and attention channels, separating model discovery from vendor disclosure.
The competitive unit in generative media may shift from a single video generator toward integrated stacks that combine models, distribution, and tools for interactive content; the reported move toward world models is consistent with that direction, though adoption remains unproven.
The trend: This is one data point in the convergence of open and semi-open model distribution, leaderboard-driven discovery, and increasingly integrated AI content-creation stacks.
🚨 Breaking News Artificial Analysis suddenly introduced an unknown model that surpassed Seedance 2.0. I am uploading few of its result in few mins [image]
Spent 2 hours on Artificial Analysis's video comparison page, and Happy Horse kept showing up. So we recorded a few matchups: - Veo 3.1 Lite vs Happy Horse - Kling v3 vs Happy Horse - SkyReel v4 vs Happy Horse - PixVerse v6 vs Happy Horse Early take: it's surprisingly good at [vi…
Prompt [2/4]: A golf ball in a cup rolling around the rim three times before finally dropping in. The golfer's body language matches each rotation. Audio: Ball rattle, exhale, plop. [video]
We've added a new pseudonymous video model to our Text to Video and Image to Video Arenas.'HappyHorse-1.0' is currently landing in the #1 spot for Text and Image to Video (No Audio) and the #2 spot for Text and Image to Video (With Audio). Further details coming soon. Example [im…
The Artificial Analysis video leaderboard upended by a mysterious, previously unknown model called “HappyHorse-1.0,” which claimed the #1 spot globally. With a massive Elo score of 1,391, this unknown model dethroned ByteDance's heavyweight Dreamina Seedance 2.0 (Elo 1,355), [ima…
A new video model dropped at #1 on the leaderboard 👀 It's called HappyHorse-1.0, and it's currently leading in both text-to-video and image-to-video. From my testing, it's particularly good at multi-shot videos and following detailed directions 👇 [video]
Prompt [3/4]: A cat staring at its own reflection in a toaster, paw tapping the chrome surface. The distorted cat reflection taps back. Audio: Paw taps, confused meow. [video]
Some unknown model surpassed the well received Seedance 2.0 on Artificial Analysis benchmark. Really curious and excited to see how it performs and what company developed it.
Unfortunately, the ranking is so misleading, all the videos are so bad, but someone keeps voting positively for the model. We are at the point where people do live benchmark manipulation.
Example generations from HappyHorse-1.0 compared to Dreamina Seedance 2.0, Kling 3.0 Pro, grok-video-imagine and PixVerse V6 (Text to Video with Audio): Prompt [1/4]: A hula hoop spinning on a kid's waist, gradually climbing to their chest, then dropping to knees, then [video]