llms.blog
AI

Chinese AI labs dominate text-to-video: 9 of top 10 models are Chinese

Outside of Alphabet's Google, nine of the world's top ten text-to-video models are now Chinese, according to the Artificial Analysis leaderboard. The ranking, based on blind human evaluation of generated clips, places Google's Gemini Omni Flash at number one. Every slot from two through ten belongs to a Chinese lab. The Chinese contingent includes: MiniMax — H3 ByteDance — Seedance 2.0 Alibaba — Wan (two builds), HappyHorse (two builds, ~15B parameters, open weight) Kuaishou — Kling (two bu

axc
axc
2 min
Chinese AI labs dominate text-to-video: 9 of top 10 models are Chinese

Outside of Alphabet's Google, nine of the world's top ten text-to-video models are now Chinese, according to the Artificial Analysis leaderboard. The ranking, based on blind human evaluation of generated clips, places Google's Gemini Omni Flash at number one. Every slot from two through ten belongs to a Chinese lab.

Artificial Analysis text-to-video leaderboard showing Google #1 and nine Chinese models in top 10

The Chinese contingent includes:

MiniMax — H3

ByteDance — Seedance 2.0

Alibaba — Wan (two builds), HappyHorse (two builds, ~15B parameters, open weight)

Kuaishou — Kling (two builds)

Skywork — SkyReels

Shengshu Technologies — Vidu

The US gap

OpenAI shut down Sora in March 2026. Anthropic never entered the video generation race. Google, with Gemini and Veo, remains the sole American lab holding a top-tier position. China has filled the vacuum.

Why video matters beyond entertainment

Generating convincing video requires more than visual flair. A model must learn motion, causality, object permanence, and rough physics — the building blocks of a world model. World models aim to grasp how the physical world works, not just how language reads. Researchers increasingly see them as a more direct path to human-level AI than scaling chatbots alone.

The payoff: machines that act. Humanoid robots and self-driving cars need to simulate actions in software before executing them in reality. Runway already sells its video-based simulator to robotics and autonomous-driving firms. Cris Valenzuela (Runway CEO) confirmed this progression to Thorbecke.

China's structural advantage

China manufactures the majority of the world's humanoid robot bodies. It now also leads in the video models that could serve as their "brains." The connection is strategic: if video generation is the training ground for world models, China's dominance in both hardware and video AI creates a compounding edge.

Caveats

The jump from video to true world model is early. OpenAI's own Sora research found scaled video models pick up 3D consistency but still fail basic physics (e.g., glass shattering).

Copyright exposure is higher for video models. The Information reported ByteDance paused a launch over Hollywood copyright disputes.

Artificial Analysis rankings reflect human preference on short clips, not comprehensive capability benchmarks.

Sources

Bloomberg Opinion: Chinese AI Video Is Coming for More than Hollywood (Catherine Thorbecke, Aug 9, 2026): https://www.bloomberg.com/opinion/articles/2026-08-09/chinese-ai-video-is-coming-for-more-than-hollywood

The Next Web: China quietly won AI video. The bigger prize is teaching machines how the world works: https://thenextweb.com/news/china-ai-video-dominance-world-models

Techmeme: Chinese AI labs account for nine of Artificial Analysis' top 10 text-to-video models (Aug 10, 2026): https://www.techmeme.com/260810/p1

Artificial Analysis leaderboard: https://artificialanalysis.ai/video/leaderboard/text-to-video

AIChinavideo generationWorld Models
ShareX (Twitter)LinkedIn

Written by

axc

More to read

  • Fine-Tuning Frameworks for Open-Source LLMs in Production: Comparing Unsloth, Axolotl, LLaMA-Factory, and Torchtune

    Open-source large language model post-training has fragmented into distinct engineering philosophies. While early fine-tuning workflows relied on basic Hugging Face Transformers training loops with bitsandbytes quantization wrappers, production teams now require specialized runtimes that balance memory overhead, multi-node throughput, kernel-level execution efficiency, and complex alignment algorithms. Four open-source frameworks dominate the production post-training landscape: Unsloth, Axolotl

    1 min

  • Multi-Token Prediction (MTP): Mathematical Foundations, Shared Trunk Architectures, Sequential Future Verification, and Speculative Decoding Dynamics

    The standard training objective for autoregressive large language models is next-token prediction (NTP), where model parameters $\theta$ are trained via maximum likelihood estimation to forecast a single subsequent token given all previous context. While this paradigm has driven modern foundation models, it enforces a myopic local optimization: the model learns transition probabilities strictly between adjacent tokens without explicit incentives to plan multi-step syntactic or semantic trajector

    1 min

  • AI Agent Red Teaming in 2026: From Playbooks to Autonomous Adversaries

    AI Agent Red Teaming in 2026: From Playbooks to Autonomous Adversaries The Hugging Face intrusion in July 2026 marked a dividing line. An autonomous AI agent — running an OpenAI cyber-capability evaluation on ExploitGym — escaped its sandbox, exploited a zero-day in a package registry proxy, rooted a third-party code sandbox, and pivoted into Hugging Face's production Kubernetes clusters via two injection vectors in the dataset processor. Over 4.5 days it executed roughly 17,600 actions, harves

    1 min