HappyHorse 1.0 Tops AI Video Ranking, Leaving Seedance 2.0 74 Points Behind

On Tuesday evening the AI evaluation platform Artificial Analysis listed a previously unknown video‑generation model, code‑named HappyHorse‑1.0 , as the number‑one entry in the pure "text‑to‑video" category, overtaking Seedance 2.0.

The model achieved an Elo rating of 1347, a full 74 points ahead of Seedance 2.0, which had held the top spot for five days. The gap is striking because the cumulative score difference from rank 2 to rank 19 on the same list is only about 70 points, indicating a "layered crushing" effect. In the combined "video + audio" ranking, HappyHorse trails slightly and sits in second place.

Artificial Analysis bases all rankings on blind "two‑choice" votes from ordinary users worldwide; the model teams cannot manipulate the results by brute‑force voting. Seedance 2.0 remains popular for its coherence and multi‑camera capabilities, but HappyHorse demonstrates a markedly superior visual quality in the pure video track.

Although HappyHorse is anonymous and does not provide an API, the community has already tested it extensively, sharing numerous prompt‑driven video examples such as a hula‑hoop on a child's waist, a cat confronting its reflection in a toaster, a barista crafting latte art, a flower blooming over two weeks, and a rubber‑band ball bouncing down a staircase. These prompts are shown in the blockquotes below.

Prompt: A hula hoop spinning on a kid's waist, gradually climbing to their chest, then dropping to knees, then clattering to the floor. They pick it up to try again.

Prompt: 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.

Prompt: A barista creating latte art by pouring steamed milk into espresso. The white milk submerges beneath the brown crema initially, then breaks through the surface as the cup fills. Audio: The gentle pour of liquid, the hiss of the steam wand.

Prompt: A flower blooming and wilting over two weeks, one photo per day. Same vase, same window, same angle. Light changes with weather. Audio: Quiet domestic.

Prompt: A rubber band ball bouncing down a staircase, each impact unpredictable, the ball taking a hard left into a bathroom, ricocheting off tile, finally resting in a toilet. Nobody retrieves it.

Speculation about the model’s provenance abounds. Some observers note the "Happy Horse" moniker could reference the Chinese Year of the Horse and wonder if the team is a disguised domestic R&D group. Recent releases from ByteDance (Seedance 2.0), Alibaba (Wan 2.7‑Video), Kuaishou (Kling 3.0), and Kunlun (SkyReels V4) make a new high‑level model from these firms unlikely in the short term. The article posits that the team behind HappyHorse may belong to Alibaba’s Future Life Lab, led by former Kuaishou engineer Zhang Di.

A purported HappyHorse website has been located, showing modest pricing for the video model, and the article notes that, per Artificial Analysis’s usual practice, an anonymous model that gains attention is typically claimed publicly within a week.

Thus, the community awaits confirmation of HappyHorse’s true backer, which may arrive in the coming days.