Supporters of Marcus Endicott’s Patreon can access weekly or monthly video consultations on this topic.
Taken together, these systems represent China’s core cohort of globally oriented large language models, led by Alibaba and Baidu embedding LLMs deeply into cloud, search, and enterprise ecosystems, while newer players such as DeepSeek emphasize efficient reasoning models with public access, ByteDance focuses on mass-market consumer deployment through its social and content platforms, and Zhipu AI positions its GLM series as a bilingual, research-driven alternative aligned with international LLM capabilities, collectively illustrating a strategy that combines large-scale domestic deployment, growing openness, and competitive parity with leading global AI models.
ModelScope (魔搭) is an AI open-source community and innovation platform founded in June 2022 by Alibaba’s Tongyi Lab (Institute for Intelligent Computing) together with the China Computer Federation Open Source Development Committee, positioned around “model-as-a-service” and providing a consolidated hub for hosting and discovering models, datasets, and “Studio” applications, alongside documentation, community features, and an “MCP Plaza” and AIGC area; it organizes content across major AI domains including NLP, computer vision, speech, multimodal, and scientific computing, surfaces frequently updated trending models and datasets (with visible dates, views, and engagement), and offers an end-to-end tooling stack—such as a Python library for inference/training/evaluation/export plus frameworks for evaluation (Eval-Scope), training/inference (Swift), and agent building (ModelScope-Agent); it also promotes low-friction development via browser-based notebooks and limited free CPU/GPU resources tied to an Alibaba Cloud account, supports partner ecosystems, and maintains governance and compliance materials (codes of conduct, user agreements, privacy, and AI ethics), aiming to sustain a collaborative Chinese AI open-source ecosystem through shared resources and shared experimentation.
[6x Feb 2026]
Alibaba Qwen series – Large language model family developed by Alibaba, deployed through Tongyi Qianwen as a public-facing web chatbot and enterprise AI platform at tongyi.aliyun.com.
Baidu Ernie / Ernie Assistant – Proprietary LLM system from Baidu, offered as a public-facing conversational assistant integrated into Baidu’s ecosystem and accessible at yiyan.baidu.com.
DeepSeek (R1 / V3) – General-purpose and reasoning-oriented LLMs developed by DeepSeek, available to the public via a standalone chatbot interface at chat.deepseek.com.
Doubao 1.5 Pro – Consumer-facing large language model created by ByteDance, delivered as a public web-based AI assistant through doubao.com.
GLM-4 and GLM-4.5 (ChatGLM) – Bilingual conversational and reasoning models developed by Zhipu AI, accessible through the ChatGLM public chatbot interface at chatglm.cn.
Kimi AI with K2.5 – Kimi is an advanced AI chatbot and large language model family developed by Moonshot AI, a Beijing-based Chinese startup founded in 2023 by Tsinghua alumni Yang Zhilin, Zhou Xinyu, and Wu Yuxin.
(In testing the first five systems listed here in February 2026, only Qwen could correctly answer "What is 'China Meta Guide'?". Doubao access was blocked by region restriction in Australia. Kimi was added later.)