智谱三大类模型开源,北京如何布局“开源之都”
Bei Ke Cai Jing·2025-04-19 13:35

Core Viewpoint - The open-source community in Silicon Valley is experiencing a global wave of AI-generated content, exemplified by the creation of an AI short film based on "Tom and Jerry," which has garnered millions of views online [1]. Group 1: AI Development and Open Source - The AI technology behind the viral content is derived from Beijing-based model company Zhipu's open-source video model CogVideoX-5B, which has sparked significant interest in the industry [2]. - Zhipu has announced 2025 as its open-source year, recently releasing a series of GLM models under a permissive MIT license, allowing free commercial use and distribution [2]. - Zhipu's president demonstrated the AutoGLM model, which can perform tasks while engaging in deep thought, showcasing advancements in AI capabilities [2][6]. Group 2: Funding and Investment - Beijing's AI Industry Investment Fund has committed an additional 200 million RMB to Zhipu, bringing the total funding for the year to over 2 billion RMB, with all rounds of financing supported by state-owned enterprises [3]. - Zhipu has completed 16 rounds of financing since its establishment in 2019, attracting investments from various state-owned enterprises across multiple cities [13]. - The company is the first among the "AI Six Tigers" to initiate the IPO process, having filed for guidance with the Beijing Securities Regulatory Bureau [16]. Group 3: Model Capabilities and Performance - Zhipu's models are categorized into three stages: search, reflection, and rumination, with the latter capable of generating detailed reports in a significantly shorter time than human counterparts [4][5]. - The GLM-Z1 model can process up to 200 tokens per second, and its cost-effectiveness is highlighted as being one-thirtieth of the Deepseek-R1 model [8]. - Over 6,000 enterprises and developers have accessed Zhipu's latest models, with token usage increasing by over 100% daily [9]. Group 4: Community and Ecosystem Development - Zhipu has announced a 300 million RMB fund to support global AI open-source community development, encouraging entrepreneurial projects based on open-source models [11]. - Beijing aims to establish a robust open-source ecosystem, with plans to organize the development of RISC-V processors and large models to attract more tech companies and developers [12].

智谱三大类模型开源,北京如何布局“开源之都” - Reportify