Chinese tech giants are making significant strides in artificial intelligence, challenging Western dominance with advancements in robotics and video generation. This week saw Alibaba, ByteDance, and Kuaishou unveil new AI models, signaling a competitive landscape where Chinese firms are rapidly closing the gap with their U.S. counterparts.
This development aligns with sentiments expressed by Google DeepMind CEO Demis Hassabis, who previously told CNBC that Chinese AI models were only “months behind” Western rivals. The latest releases demonstrate China’s growing prowess in AI, directly competing with leading models like OpenAI’s Sora for video generation and those from Nvidia and Google in the robotics space.
Alibaba’s DAMO Academy introduced RynnBrain, an AI model designed to enhance robots’ understanding of their physical environment and object recognition. A demonstration showcased a robot successfully identifying and manipulating objects like oranges, and retrieving items from a refrigerator. This capability is crucial for robotics, as even seemingly simple tasks require extensive training for AI to accurately identify and interact with everyday objects. RynnBrain positions Alibaba as a competitor to Nvidia and Google in the development of AI for robotics.
A key innovation in RynnBrain is its integrated time and space awareness. This allows robots not just to react to immediate stimuli but also to recall past events, track task progress, and execute multi-step operations, leading to more reliable performance in complex real-world scenarios. Alibaba’s broader objective appears to be establishing a foundational intelligence layer for embodied systems.
ByteDance has launched Seedance 2.0, an AI video generation model that creates realistic videos from text prompts, and can also incorporate existing videos and images. Early reviews suggest Seedance 2.0 produces highly realistic AI-generated visuals. Billy Boman, who runs an AI content creation agency, noted the rapid evolution of AI video generation, stating that capabilities have drastically improved from the limitations of earlier models. Yakefu from Hugging Face highlighted Seedance 2.0’s advancements in controllability, speed, and production efficiency, calling it one of the most well-rounded video generation models tested.
However, Seedance has faced controversy. Reports indicate that a feature allowing AI voice generation from uploaded images has been suspended following concerns about consent. ByteDance has not yet provided a comment on this matter.
Kuaishou’s Kling 3.0 is another contender in the AI video generation space, offering significant upgrades in consistency, photorealistic output, extended video durations of up to 15 seconds, and native audio generation across multiple languages and accents. While currently available only to paying subscribers, Kuaishou plans to make it publicly accessible soon. The success of its Kling models has been a notable driver for Kuaishou’s stock performance, with shares rising over 50% in the past year.
Beyond these major players, other Chinese AI companies are also making waves. Zhipu AI, trading as Knowledge Atlas Technology, saw its shares surge following the release of GLM-5, an open-source large language model with advanced coding and agent task capabilities. The company claims GLM-5 rivals Anthropic’s Claude Opus 4.5 in coding benchmarks and surpasses Google’s Gemini 3 Pro on certain tests. Additionally, MiniMax’s stock climbed after launching its updated M2.5 open-source model, which includes enhanced AI agent tools designed to automate tasks.
Original article, Author: Tobias. If you wish to reprint this article, please indicate the source:https://aicnbc.com/18526.html
