The opening ceremony of the 2026 Youth Olympics in Dakar needed a drum performance with many faces held consistent across a continuous transition. Staging and choreographing such a large performance was impractical. Alibaba's answer was to synthesize the sequence with its Wan video model — and that same engine is now sitting at the top of the two benchmark charts creators actually watch.
Alibaba's video generation has had a quiet, relentless 2026. Wan3.0, the company's current flagship, holds the number-one position on Artificial Analysis' text-to-video and video-editing leaderboards simultaneously — a double crown that, per Alibaba's own statements at the Apsara Conference (云栖大会) in September, it pairs with a next-generation model scheduled to launch in November 2026.
Key takeaways
- Wan3.0 ranks #1 on Artificial Analysis' text-to-video and video-editing leaderboards, according to Alibaba's Apsara Conference disclosures carried by Xinhua Finance.
- The model generates single 30-second clips with native audio-video sync and accepts structured inputs like documents, spreadsheets, and web pages.
- Alibaba stated a next-generation video model will launch in November 2026, emphasizing longer duration, tighter control, and "director-level" narrative understanding.
- The Wan engine was used at the Dakar 2026 Youth Olympics opening to synthesize an AI-made drum performance and visitor self-videos.
Why the double leaderboard matters
Artificial Analysis is the benchmark most Western creators and procurement teams cite, so a simultaneous #1 on both its video charts is a credibility signal that travels. Alibaba says Wan3.0 already produces 30-second clips with native audio sync and can take structured inputs — a document or a webpage — rather than only a text prompt. That shifts the model from "describe a scene" toward "turn my asset into a scene," which is closer to how studios actually work.
The next-generation model, Alibaba says, pushes further on the harder problems: keeping a character, setting, and camera coherent over longer spans, and understanding full narrative structure instead of stitching single shots. If it ships in November as stated, it will land straight into a crowded field where rivals such as Kuaishou's Kling and ByteDance's Seedance are described as pursuing longer, more controllable output as well.
From cloud backbone to on-screen content
The Dakar Youth Olympics, which opened 31 October 2026, is described in Chinese event coverage as the first Olympic event held on the African continent. Alibaba Cloud was the cloud and AI provider, and — unusually — the AI became something spectators could touch. At the opening, an AI-made drum performance fused performers' footage into one continuous transition using the Wan model, with each person's face, position, costume, and drum kept consistent across the cut. On the ground, a station let visitors create a short video of themselves generated with Wan.
This is a useful window into how Chinese cloud providers monetize generative media: not only as APIs, but as bundled experience at flagship events where the brand demonstration is the marketing. Alibaba has now supported multiple Olympic Games (Tokyo 2020, Beijing 2022, Paris 2024, Milan-Cortina 2026, Aichi-Nagoya 2026, and Dakar 2026), and it leaves the local cloud infrastructure in place afterward as national-capability projects.
The honest competitive picture
Calling any single model "the best" is a moving target. Benchmark leadership reflects one evaluator's methodology at one moment, and rivals contest the same territory weekly. What is unambiguous is that Alibaba has chosen to compete on length, control, and production integration rather than on a single eye-catching clip. The November launch may be the real test of whether Wan's lead holds once competitors' next versions land.
Honest limitations
This article's core claims — Wan3.0's dual #1 leaderboard positions and the November 2026 next-generation launch — are drawn from Alibaba's Apsara Conference disclosures as reported by Xinhua Finance (a state financial wire, treated as authoritative), and corroborated by Alibaba Cloud's own model catalog noting Wan3.0's 30-second, multimodal-reference, audio-sync capabilities. The Dakar Olympics usage is reported in Chinese event coverage; production details are presented as reported at the event and were not independently verified against a primary technical paper. Leaderboard rankings are a single evaluator's snapshot and can shift. I have not run Wan3.0 myself, have not benchmarked it against Kling or Seedance, and have not confirmed November pricing or feature scope of the upcoming model.
What readers can do now
- If you choose a video model, check the specific leaderboard, not the headline. Text-to-video and video-editing are different skills; a dual #1 is rarer than a single top score.
- Match model to workflow. If you feed documents or web pages into prompts, verify the candidate actually accepts structured inputs before betting a pipeline on it.
- Watch November. Alibaba's next model launch is a date worth tracking if you are comparing Chinese video engines against Runway, Kling, or Seedance for a 2027 production plan.