For scale: Anthropic alone (~$965B, May 28, 2026) is worth about 4.6× every tracked Chinese lab combined (~$209B). via CNBC
Tsinghua-origin GLM stack for enterprise, agents, video, and state-linked deployments.
Open reasoning models; cost discipline; efficient training.
$45-50B reported first external raise talks
Kimi consumer assistant; long context; open model releases.
$2B round at $20B valuation
MSRA-heavy Shanghai lab; multimodal models and heavy training infrastructure, increasingly oriented toward on-device and edge deployment.
$10B reported IPO target valuation
Former general LLM startup now oriented around healthcare and medical models.
~$2.8B last disclosed valuation; now pivoted to healthcare AI
Kai-Fu Lee's startup that left the frontier-pretraining race; its core team was absorbed into Alibaba, leaving an enterprise/archive footprint.
$1B+ peak valuation before frontier pre-training exit
Tencent's foundation-model stack for WeChat/cloud/games, plus video and 3D.
Tencent-owned team; no standalone lab valuation used
ByteDance's Doubao app, Seed research, video models, Volcano Engine calls, and on-device agent experiments.
ByteDance-owned team; no standalone lab valuation used
Alibaba Cloud's open model family, Bailian MaaS layer, coding/agent benchmarks, and developer distribution.
Alibaba-owned team; no standalone lab valuation used
Xiaomi AI Lab's open models tied to device, car, and edge-product ambitions.
Xiaomi-owned team; no standalone lab valuation used
Efficient MoE models for agent/tool-use and local-services workflows.
Meituan-owned team; no standalone lab valuation used