智谱开源新一代旗舰模型GLM-4.7：超越GPT-5.2，正冲刺「全球大模型第一股」

Core Viewpoint - The article discusses the launch of GLM-4.7 by Zhipu AI, which aims to become the "first global large model stock" and highlights its advancements in coding capabilities, reasoning, and tool collaboration [2][11]. Group 1: Model Performance - GLM-4.7 has achieved the best performance among open-source models in multiple mainstream public benchmark tests [6]. - In the Code Arena, a global coding assessment system, GLM-4.7 ranks first among open-source models and first among domestic models, surpassing GPT-5.2 [8]. - The model shows significant improvements in various coding benchmarks, achieving 73.8% in SWE-bench-Verified, 84.9% in LiveCodeBench V6, 66.7% in SWE-bench Multilingual (an increase of 12.9%), and 41% in Terminal Bench 2.0 (an increase of 16.5%) [8]. Group 2: Technical Enhancements - GLM-4.7 has enhanced programming capabilities, showing improved performance in multi-language coding and terminal agents, with a "think first, act later" mechanism in various programming frameworks [3]. - The model has made notable advancements in front-end generation quality, producing visually appealing web pages, PPTs, and posters [4]. - Tool invocation capabilities have been improved, scoring 67.5 in BrowseComp and achieving 87.4 in τ²-Bench, surpassing Claude Sonnet 4.5 [4]. Group 3: Revenue and Growth - Zhipu AI is projected to be the largest independent large model vendor in China by 2024, with revenues of 57 million yuan in 2022, 125 million yuan in 2023, and 312 million yuan in 2024, reflecting a compound annual growth rate of 130% [11]. - In the first half of 2025, Zhipu AI's revenue reached 190 million yuan, showing a year-on-year growth of over 300% [11]. - The company has completed eight rounds of financing, raising over 8.3 billion yuan, with investments from major firms and government support [11].