HappyHorse 1.1: #1 AI Video Generator from Alibaba

A mysterious new contender has just seized the crown in AI video generation. HappyHorse-1.0 — a pseudonymous video model competing under a playful codename — has stormed to the top of the Artificial Analysis Video Arena leaderboard with an astonishing ELO of 1,332, leaving the second-place Dreamina Seedance 2.0 trailing by a massive 59 points. It's one of the largest gaps ever recorded on the leaderboard.

What Is HappyHorse-1.0? Alibaba's New AI Video Model Explained

On April 8, 2026, Artificial Analysis — the authoritative AI benchmarking platform — added a new pseudonymous model to its Text to Video and Image to Video Arenas: HappyHorse-1.0. The name might sound whimsical, but its performance is anything but.

According to Artificial Analysis's official announcement, HappyHorse-1.0 currently holds:

• #1 on the Text to Video (No Audio) Leaderboard — ELO 1,332 with 2,784 evaluation samples, leading by 59 points
• #2 on the Text to Video (With Audio) Leaderboard — second only to Dreamina Seedance 2.0, which has inherent audio generation advantages
• #1 on the Image to Video (No Audio) Leaderboard — equally dominant in image-to-video generation, ranked Top 1
• #2 on the Image to Video (With Audio) Leaderboard — second only to Dreamina Seedance 2.0, which has inherent audio generation advantages

Best AI Video Models 2026: HappyHorse vs Sora 2 vs Kling 3.0 vs Seedance 2.0

To understand just how dominant HappyHorse-1.0 is, let's look at the top of the Artificial Analysis Text to Video (No Audio) leaderboard:

What does a 59-point ELO gap mean? In AI model blind tests, an ELO difference over 50 is generally considered a "clear advantage," and anything over 100 is essentially a blowout.

HappyHorse-1.0 Review: How Good Is It Really?

Artificial Analysis showcased multiple generation samples from HappyHorse-1.0 in their announcement. Based on available information, it excels in several key areas:

• Motion consistency — Characters and objects move fluidly and naturally, without the common flickering and deformation artifacts
• Physical realism — Water surfaces, smoke, fabrics, and other complex physical effects reach new heights
• Long-take stability — Maintains visual consistency throughout longer continuous shots, an area where many previous models struggled
• Prompt adherence — Understanding of complex prompts is noticeably better than competitors, with generated visuals closely matching descriptions

HappyHorse 1.1 is now available in Oimi Canvas for text-to-video, first/last-frame image-to-video, and reference-image video workflows, so creators can test the model directly instead of only comparing blind Arena samples.

Who Made HappyHorse-1.0? It's From Alibaba

This is the question everyone in the community is asking. HappyHorse-1.0 entered the Arena as a "pseudonymous model," meaning its developer's identity is temporarily hidden.

Anonymous testing isn't unusual in the AI industry. OpenAI previously tested GPT-Image-2 under codenames like maskingtape-alpha on the LMSYS Arena, and Google has done the same with Gemini models. HappyHorse's sudden dominance has sparked intense speculation:

• Is it a secret project from a tech giant — Google, Meta, or Anthropic?
• A breakthrough from an ambitious startup?
• Or perhaps a surprise from the open-source community?

Artificial Analysis has stated that "further details are coming soon,"

but clues on social media are already converging. According to cross-verification by multiple tech bloggers, HappyHorse-1.0 is almost certainly from Alibaba. If true, it wouldn't be entirely surprising — Alibaba previously released Wan 2.6 (ranked #23 on the leaderboard), and HappyHorse-1.0 likely represents a major leap in their video generation capabilities. With an ELO 177 points higher than Wan 2.6, this isn't just an incremental improvement — it looks like an architectural breakthrough.

Best AI Video Generators in 2026: Complete Comparison

Looking back at Q1 2026, the AI video landscape has become fiercely competitive. In just the last month alone, three major new models entered the arena:

• HappyHorse-1.0 — the anonymous king of the leaderboard
• Dreamina Seedance 2.0 — ByteDance's new flagship, ranking second at 720p
• PixVerse V6 — the latest iteration from PixVerse

Meanwhile, Kuaishou's Kling 3.0 series, Google Veo 3, xAI's grok-imagine-video, and Runway Gen-4.5 are all battling for position. Notably, the former benchmark-setter OpenAI Sora 2 Pro has slipped to #20 on the leaderboard — a testament to how fast this industry is moving.

Which AI Video Generator Should You Use? Free & Paid Options

If you're a content creator, marketer, or simply someone interested in AI video, here's what these developments mean:

• AI video quality has reached professional grade — Videos from the top models are increasingly difficult to distinguish from real footage
• More choices at lower prices — Competition is driving prices down, ranging from $2.40/min to $30/min
• No technical background required — Most platforms offer intuitive text-to-video interfaces