Alibaba, ByteDance, and Kuaishou's latest AI models have significantly narrowed the gap with their American competitors.

Alibaba, ByteDance, and Kuaishou's Latest AI Models Narrow the Gap with US Competitors

While the US discussed AI in finance and programming, Chinese giants unveiled new models for robotics and video generation, challenging claims of a lag.

Alibaba DAMO Research introduced RynnBrain—a model that teaches robots to understand the physical world. In a demonstration, a mechanical claw was shown picking oranges and retrieving milk from a refrigerator. A key feature is its built-in awareness of time and space: the robot remembers event sequences and performs multi-step tasks, bringing its behavior closer to that of humans. Alibaba now enters direct competition with Nvidia and Google in the robotics field.

ByteDance launched Seedance 2.0 for generating realistic videos from text descriptions. Experts call it one of the most balanced models in terms of controllability, speed, and quality. However, it wasn't without issues: the photo-to-voice generation feature was temporarily disabled after bloggers highlighted the risk of using others' images without consent.

Kuaishou responded with an update to Kling 3.0—a competitor to Seedance. The algorithm produces 15-second clips with photorealistic visuals and built-in audio generation in multiple languages and dialects. For now, the model is only available via subscription, but the company promises to expand access. The success of the Kling series has already boosted Kuaishou's stock by 50% over the past year.

Other players also made moves. Zhipu AI released GLM-5—an open-source language model approaching the capabilities of Claude Opus 4.5 and surpassing Gemini 3 Pro in some benchmarks. MiniMax updated its M2.5 with enhanced AI agents for task automation. Stocks of both companies reacted positively.