AI Foundations · The 2027 model lineup
You can name ten Chinese labs actively shipping at frontier pace in 2026, and explain why "DeepSeek vs everyone else" is the wrong frame for understanding China's AI position.
Refreshed: May 2026.
The default Western frame for China's AI position in 2026 is: there's DeepSeek, and there's everyone else. The frame is wrong. DeepSeek is the easiest name to remember because DeepSeek-V4 is open, high-profile, and clearly documented in English. But the public model surfaces around Qwen, Kimi, Doubao/Seed, GLM, MiniMax, StepFun, MiMo, Baichuan, and Yi show a wider field. This chapter walks that ten-lab map.
Why the "DeepSeek only" frame is wrong. Three reasons.
First, the public surfaces are not redundant. Qwen3.6-Plus is an Alibaba Cloud and Qwen App model; Kimi is a Moonshot model platform; Seed2.0 is ByteDance's Doubao/Seed model line; GLM, MiniMax, StepFun, MiMo, Baichuan, and Yi each publish a distinct model or product surface. If you collapse those into "DeepSeek," you lose the product and deployment map.
Second, the chapter's "ten" is a practical literacy roster, not a claim that no other Chinese lab matters. It gives learners enough resolution to parse frontier news without pretending every model release comes from the same place.
Chapter contains 4 lessons.