Key takeaways
A Chinese LLM, also searched as China LLM or Chinese LLMs, is a large language model trained and shipped by a company based in China. In 2026 the names global teams actually shortlist are DeepSeek, Qwen (Alibaba), GLM (Zhipu), and Kimi (Moonshot). The label is about origin and vendor, not a promise that one model is best at every job.
What the term includes
Use “Chinese LLM” for the model family and the API you can call: DeepSeek-V4-Pro, Qwen3-Max, GLM-4.7, Kimi K2.6, and the other rows in the Swift Horse catalog. It also covers open-weight checkpoints some of those labs publish, and hosted APIs on the vendor console or a cloud marketplace. It does not mean every model that can speak Chinese. GPT, Claude, and Gemini can answer in Chinese and are not Chinese LLMs.
A useful citation names four fields: vendor, model ID, the job you tested, and whether the spec is published. On this index, DeepSeek-V4-Pro lists 1.6T total parameters with about 49B active and a 1M-token context. Qwen3-Max, GLM-4.7, and Kimi K2.6 do not publish parameter count or context length here. Say “not disclosed” instead of copying an old blog number.
How to choose after the definition
Coding and hard reasoning: start with DeepSeek-V4-Pro. Multimodal work on an Alibaba stack: Qwen3-Max. Enterprise agents and tools: GLM-4.7. Long jobs with many agents: Kimi K2.6. The comparison article is /en/articles/best-chinese-llm-2026. Model pages live at /en/models/{id}.
FAQ
What is a Chinese LLM?
A large language model from a China-based lab. The 2026 shortlist on this index is DeepSeek, Qwen, GLM, and Kimi.
Is China LLM the same as Chinese LLM?
Yes. Both mean models from Chinese vendors, not merely models that reply in Chinese.
Which Chinese LLM should I try first?
Match the job. Coding: DeepSeek-V4-Pro. Multimodal: Qwen3-Max. Agents: GLM-4.7. Many agents or long workflows: Kimi K2.6.
Is this an official definition?
No. Swift Horse is an independent index. Confirm model IDs on the vendor console.