Original Research

Independently researched analysis of China's AI ecosystem. Facts, analysis, interpretation and uncertainty are clearly separated; every article includes a China AI Hub View.

• Benchmark Methodology Divergence Across Chinese Model Vendors

  Research question: To what extent are headline benchmark scores from Chinese model vendors directly comparable, and which methodological differences explain the apparent gaps between them?

  2026-09-22 · updated 2026-09-22

• Chinese AI APIs Compared: Six Official Platforms, One Convergence and a Few Real Differences

  Research question: How do the six official Chinese AI API platforms differ in practice on authentication, protocol compatibility, model routing, regions and rate limits?

  2026-09-22 · updated 2026-09-22

• Context Windows in China's 2026 Model Market: 1M Tokens as Table Stakes and the Stratification Below It

  Research question: How do context windows and maximum output tokens distribute across Chinese frontier models in 2026, and what does the tiering reveal about the market's long-context economics?

  2026-09-22 · updated 2026-09-22

• Chinese AI Coding Models: The Agent-Led Race

  Research question: How do Chinese labs compete in the coding space — through agents, subscriptions or benchmark scores — and what do the vendor-reported numbers actually establish?

  2026-09-22 · updated 2026-09-22

• Chinese AI Model Companies Explained: The Six Labs Behind the Frontier

  Research question: Who are the organizations behind China's frontier AI models, and how do their founding histories, corporate structures and product strategies explain the shape of the model landscape in 2026?

  2026-09-22 · updated 2026-09-22

• Chinese AI Model Licensing Explained: MIT, Apache and the Three Custom Regimes

  Research question: How do the licenses on China's open-weight frontier models actually differ, and what revenue or usage thresholds change what a commercial adopter owes?

  2026-09-22 · updated 2026-09-22

• How Chinese AI Model Pricing Has Changed: Flash Tiers, Cache Discounts and the $0.15 Input Standard

  Research question: What has actually changed in Chinese AI model API pricing during 2026, and what does the current price structure across six providers reveal about the economics of the market?

  2026-09-22 · updated 2026-09-22

• How Chinese Frontier Models Use Mixture-of-Experts: Active Parameters, Linear Attention and the Economics of Sparsity

  Research question: How do Chinese frontier labs design Mixture-of-Experts architectures in 2026, and what do activation ratios and linear-attention hybrids reveal about the economics of serving these models?

  2026-09-22 · updated 2026-09-22

• GLM and Agent-Oriented AI: How Zhipu Built an Agent Stack Around Open Weights

  Research question: How does Zhipu AI's GLM ecosystem operationalize agent-oriented AI, and what does the combination of open weights, a phone agent and a coding subscription reveal about the agent market in China?

  2026-09-22 · updated 2026-09-22

• Multimodal Capability in China's 2026 Model Lineup: Vision, Video and the Computer-Use Frontier

  Research question: Which Chinese frontier models actually support vision, video and computer-use in 2026, and how does multimodal capability distribute across price tiers and vendors?

  2026-09-22 · updated 2026-09-22

• How DeepSeek Changed China's AI Market: Open Weights, Price Floors and the Standard It Set

  Research question: How has DeepSeek's model, pricing and open-source strategy reshaped the competitive baseline for China's other frontier AI labs?

  2026-09-22 · updated 2026-09-22

• Open Weights and API Models in China's 2026 Ecosystem: A Structural Analysis

  Research question: How do Chinese model vendors structure the relationship between open-weight releases and API-only flagships, and where do the two tracks diverge for adopters?

  2026-09-22 · updated 2026-09-22

• The Rise of Chinese AI Agents: What the 2026 Landscape Actually Looks Like

  Research question: How is China's AI agent landscape structured in 2026 — which companies build which agent types, and what patterns emerge from the distribution?

  2026-09-22 · updated 2026-09-22

• The State of China's AI Models in 2026: A 19-Model Panorama

  Research question: What does China's frontier model landscape look like in 2026, and what patterns emerge from the distribution of open vs closed, context length and pricing across 19 tracked models?

  2026-09-22 · updated 2026-09-22