智谱API订阅制取得显著商业化进展 目前年度经常性收入已过亿

Core Insights - The CEO of Zhiyu, Zhang Peng, announced that the company's software tools and model business (GLM coding plan) has achieved an annual recurring revenue (ARR) of over 100 million RMB (approximately 14 million USD) [1] - Despite being average compared to American competitors, this figure indicates significant progress in convincing Chinese developers to pay for AI services [1] - Zhiyu expects to see over 100% revenue growth by 2025 [1] Revenue Diversification - Zhiyu is diversifying its revenue structure by shifting focus from government and enterprise clients to embrace developers in China and other regions [1] - The company aims to prioritize model applications and API services, with a goal of increasing the revenue share from API business to 50% [1] - Currently, Zhiyu's API platform serves over 2.7 million paying customers, including some of China's largest tech companies [1] Subscription Model - The API subscription model allows developers to subscribe to services on a monthly or annual basis [1] - In September, Zhiyu launched an AI-driven coding tool subscription plan priced as low as 20 RMB per month, which is about one-seventh of the price of Anthropic's Claude [1] - The coding tool plan has reportedly attracted over 150,000 users [1] Competitive Positioning - Zhang Peng stated that Zhiyu's models are among the best in the world, emphasizing significant advantages in pricing and costs [2] - The company aims to become the first publicly listed AI large model vendor in China [2] - There is a strong demand for high-quality AI services in the Chinese market, which currently exceeds supply [2] Model Performance - Zhiyu was founded in 2019 by researchers from Tsinghua University and has received support from Alibaba, Tencent, and various local government funds [2] - The GLM-4.6 model from Zhiyu ranks just below top models from Silicon Valley on the benchmark site LMArena, alongside models from DeepSeek and Alibaba's Qwen [2] - The recent releases of GLM-4.5 and GLM-4.6 have garnered global attention for their performance, particularly in enhancing programming and intelligent agent capabilities [2]