Regional guide · reviewed 2026-08-21
AI Models Built in China: Qwen, DeepSeek and More
China has the broadest coverage in the current AIOAI index, spanning general assistants, coding models, multimodal systems and dedicated video generators.
20 current models
Use the task rankings to separate general-purpose models from specialist image and video products. Availability and pricing can differ by market, so confirm details on each official site.
01Qwen
A broad model family for reasoning, coding, multimodal work and multilingual tasks.
DeepSeek
Reasoning and coding models with a strong free experience and developer access.
Seedance
ByteDance's first-party video-generation model family for prompt adherence, motion, multi-shot storytelling and audio-visual creation.
04GLM
A multilingual model family spanning chat, reasoning, coding and visual understanding.
05Kling AI
A visual generation platform focused on cinematic image-to-video and text-to-video work.
Wan
Alibaba's video-generation model family with hosted creation tools, APIs and openly released model weights.
07Doubao
A consumer-friendly assistant with text, image and voice experiences.
08Vidu
A first-party video model and creation platform for text-to-video, image-to-video and reference-driven character consistency.
ERNIE
Baidu's multimodal foundation-model family for text, reasoning, vision, audio and enterprise applications.
Kimi
A research and long-document assistant with reasoning and coding capabilities.
11HappyHorse
Alibaba ATH's first-party image-to-video model, exposed through Model Studio for motion-rich video generation from a starting frame.
12Step
A multimodal model family covering reasoning, coding, image, video and speech generation.