China's AI Tigers Evolve | Generated by AI
China’s First-Wave AI Startups: From the “Tigers” to Today’s Leaders
You’re spot on—the inaugural batch of generative AI startups in China, often dubbed the “AI Tigers” or “Six Tigers,” kicked off around 2022–2023 amid the global LLM boom. This group includes Zhipu AI, Baichuan Intelligence, 01.AI, DeepSeek, Moonshot AI (home to the Kimi chatbot), and MiniMax. They emerged as Tsinghua and other top university spinoffs, backed by heavy state and VC funding to counter U.S. dominance. Early on, they all hit unicorn status quickly, raising billions collectively and releasing competitive models like GLM (Zhipu) and Baichuan-2.
Fast-forward to October 2025: The pack has thinned in terms of raw performance and hype. DeepSeek, Zhipu, and Moonshot (Kimi) have pulled ahead decisively, dominating benchmarks, user adoption, and innovation cycles. The others—Baichuan, 01.AI, and MiniMax—are still solid players with strong funding but lagging in frontier model quality and market buzz. DeepSeek’s open-source ethos and cost-efficiency have made it the breakout star, while Zhipu and Moonshot excel in multimodal and long-context tasks. Here’s a quick comparison based on the latest data:
| Company | Key Models (2025) | Benchmark Highlights (e.g., SuperCLUE, LMSYS Arena) | Funding/Valuation (as of Q3 2025) | Standout Edge |
|---|---|---|---|---|
| DeepSeek | DeepSeek-V3, R1, DeepSeekMath | #1 on SuperCLUE for reasoning/math/code; top performer in China, rivals GPT-4o/Claude 3.5 | ~$1B+ raised; valuation est. $2–3B (exact undisclosed) | Open-source leader; best cost-performance ratio; biggest 2025 AI story |
| Zhipu AI | GLM-4.5, GLM-4.5V (visual) | #6 overall; 64%+ on SWE-bench (beats Kimi K2); strong in multilingual tasks | $2.7B+ raised; $2B+ valuation; IPO filed April 2025 | Academic roots (Tsinghua); multimodal push (image/video gen) |
| Moonshot AI (Kimi) | K2, K1.5, Moonlight-16B | #5 overall; excels in 2M-token context; competitive vs. DeepSeek-R1 on MMLU-like tests | $2.5B+ round led by Alibaba; $3.3B valuation | Long-context mastery; consumer chatbot dominance |
| Baichuan | Baichuan-3 series | Mid-tier (behind top 3); solid but not frontier on Arena | $700M+ recent round; $2.8B+ valuation | Enterprise focus; broad VC backing |
| 01.AI | Yi series (e.g., Yi-1.5) | Lower rankings; good for general tasks but trails in reasoning | $300M+ raised; unicorn (~$1B) | Backed by ex-Baidu founder; steady but not explosive |
| MiniMax | MiniMax-Text-01, M1-80k | Competitive in long-context (1M tokens); IPO eyeing | $1B+ valuation; strong consumer apps | Video/image gen strength; transitioning to open models |
In short, yes—DeepSeek, Zhipu, and Kimi are the clear frontrunners now. DeepSeek leads in raw power and accessibility, Zhipu in ecosystem breadth (e.g., AGI ambitions via GLM), and Kimi in practical usability for Chinese users. The shift reflects faster iteration: DeepSeek’s R1 drop in early 2025 sparked a “China investible again” wave, pulling in $5B+ across the sector. Baichuan and 01.AI hold ground in enterprise/closed models, but they’re not matching the open-source hype or benchmark leaps. If trends hold, expect Zhipu to IPO first, cementing the trio’s edge.
Ranking the Chinese Open Model Builders
The 2025 China AI Index
Meet the ‘Six Tigers’ of China’s AI
How DeepSeek Made China Investible Again
Zhipu AI Releases GLM-4.5