Z.ai has released GLM-4.7, a new open-source large language model focused on software development, intensifying competition among Chinese artificial intelligence firms as several prepare for public listings. The model launched on December 26, just days after parent company Zhipu AI completed its listing hearing with the Hong Kong Stock Exchange.
According to the company, GLM-4.7 delivers leading performance among open-source models on key programming benchmarks. It scored 73.8% on SWE-bench Verified, topped Code Arena’s WebDev leaderboard, achieved 87.4% on τ²-Bench, 67.5% on BrowseComp, and 95.7% on the AIME 2025 mathematics benchmark. Z.ai said the model significantly outperforms its previous release, GLM-4.6, in real-world coding and development tasks.
The launch comes only hours after rival Chinese startup MiniMax introduced its M2.1 model, underscoring the pace of innovation in China’s AI sector. Both companies recently passed Hong Kong listing hearings, with Zhipu AI reportedly aiming to raise about $300 million. Despite the move toward public markets, Zhipu said it plans to continue contributing to open-source artificial general intelligence research in 2026.
Founded in 2019 by professors from Tsinghua University, Zhipu AI is backed by Alibaba and has raised 8.36 billion yuan across eight funding rounds, reaching a valuation of roughly 40 billion yuan. GLM-4.7 marks the company’s third major model update in recent months.
Developers can access GLM-4.7 through the BigModel.cn API, Z.ai’s own platform, and open-source model weights hosted on Hugging Face.
