Alibaba's Wan 3.0 Tops the Video Leaderboard at 1175 Elo, $0.20 per Second
Alibaba's Wan 3.0 has topped the Artificial Analysis ranking of text-to-video models that generate clips with sound: 1175 Elo as of November 13, 2026. Elo is a chess-style rating built from pairwise comparisons of videos. A second of video costs $0.20, and a minute costs $12. But the gap to third place (MiniMax H3, 1161) is just 14 points against a margin of ±13–15, so price and licensing matter as much as rank.
AI-processed from X: @ApollonVisual; edited by Hamidun News
You're choosing an AI model to make video for your product, and you look at the leaderboard. Alibaba's Wan 3.0 ranks first there: 1175 Elo in the text-to-video-with-sound table as of November 13, 2026, with a second of footage costing $0.20. Elo is a rating system borrowed from chess: models are compared head to head, and beating a strong opponent earns more points. It sounds like a clear win, but in practice first place, price and license terms are three separate things, and the headline alone doesn't reveal them.
What happened with Wan 3.0
Wan 3.0 launched on August 24, 2026, and climbed to the top of the Artificial Analysis ranking in under three months. The model generates clips up to 30 seconds long at 1080p and charges $0.20 per second. The math is simple: one 30-second clip costs $6, and a minute of video costs $12.
The scale needs a reference point. The ranking is anchored to Kling 3.0 at 1000 points. Against that baseline, Google Veo 3.1 at 964 Elo in the text-to-video task sits below the anchor, meaning it loses to it on average.
The leader isn't alone, and the price gap is large
In the adjacent image-to-video table, ByteDance's Dreamina Seedance 2.5 takes first place with 1161 Elo at a listed price of $0.569 per second. That works out to about $34 per minute of video, nearly three times the cost of Wan 3.0.
In third place for text-to-video is MiniMax H3, with the same 1161 Elo and a price of $0.08 per second, or $4.80 per minute. The model has open weights, but there's a catch. The license blocks local deployment in the US, EU, UK and South Korea, and the open weights are limited to 768p. For an American or European team, "open model" turns into a question for the lawyers.
Even cheaper are HeyGen Video 1, at 1152 Elo for $0.04 per second ($2.40 per minute), and Pika H3 Flex, at 1124 Elo for $0.035 per second ($2.10 per minute). They trail the leader by 23 and 51 points, but cost five and nearly six times less.
How to read these numbers
The main caveat is statistical. For most models in the top ten, the confidence interval is roughly ±13–15 Elo, and models within 25 points of each other are considered statistically close. Think of two thermometers reading 21.5 and 21.6 degrees: you wouldn't feel the difference. The same applies here. Wan 3.0 leads MiniMax H3 by 14 points and HeyGen Video 1 by 23, so declaring a clear winner isn't justified. The order in a group like this can flip with the next data snapshot. A meaningful gap appears only against Pika H3 Flex, where the difference exceeds 50 points.
There are other limits. The ranking reflects average preferences in head-to-head video comparisons, not your specific task. An ad spot, character animation and an educational video may each produce a different model order. The prices in the post are list prices, and your actual bill depends on the provider, your volume, and the number of failed generations, which you also pay for. The source is a social media summary, so the key figures should be checked on the Artificial Analysis page itself.
What to do in practice
- Don't pick a model based on first place alone: if the gap is under 25 points, compare price, clip length and audio on your own prompts.
- For high-volume production of short clips, test HeyGen Video 1 or Pika H3 Flex: a small quality loss may not justify paying five times more.
- If you need a long 1080p clip, Wan 3.0 with its 30-second clips is a strong candidate.
- Before planning a local run of open weights, read the license first: MiniMax H3's terms close off several key markets.
Quality among the top video models is converging, and the market is no longer about a single leader. The winner is whoever balances price, license and clip length against their own needs.
- Nov 21, 2026A swimming team captain handed ChatGPT three scheduled tasks and cut her routine work by a multiple
- Nov 21, 2026MIRA Neural Network Renders Rocket League at 20 FPS, but Ping Hits 460 ms
- Nov 21, 2026Google built a closed copy of the internet with fake banks to train Chrome's AI agents
- Nov 21, 2026MLPerf Training Adds Nvidia Rubin and TPU Ironwood in First Appearance: 153 Systems
Key Questions Answered
What happened with Wan 3.0?
Wan 3.0 launched on August 24, 2026, and climbed to the top of the Artificial Analysis ranking in under three months. The model generates clips up to 30 seconds long at 1080p and charges $0.20 per second. The math is simple: one 30-second clip costs $6, and a minute of video costs $12. The scale needs a reference point. The ranking is anchored to Kling 3.0 at 1000 points. Against that baseline, Google Veo 3.1 at 964 Elo in the text-to-video task sits below the anchor, meaning it loses to it on average.
What are the key details on The leader isn't alone, and the price gap is large?
In the adjacent image-to-video table, ByteDance's Dreamina Seedance 2.5 takes first place with 1161 Elo at a listed price of $0.569 per second. That works out to about $34 per minute of video, nearly three times the cost of Wan 3.0. In third place for text-to-video is MiniMax H3, with the same 1161 Elo and a price of $0.08 per second, or $4.80 per minute. The model has open weights, but there's a catch. The license blocks local deployment in the US, EU, UK and South Korea, and the open weights are limited to 768p.
How to read these numbers?
The main caveat is statistical. For most models in the top ten, the confidence interval is roughly ±13–15 Elo, and models within 25 points of each other are considered statistically close. Think of two thermometers reading 21.5 and 21.6 degrees: you wouldn't feel the difference. The same applies here. Wan 3.0 leads MiniMax H3 by 14 points and HeyGen Video 1 by 23, so declaring a clear winner isn't justified. The order in a group like this can flip with the next data snapshot.
Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews
Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.