GLM-4.7

GLM-4.7 is a foundation model engineered to serve as a robust partner for intelligent agents and complex development environments. With a architecture featuring 355 billion total parameters and 32 billion active parameters, the model is designed to handle lengthy task cycles that require high stability and consistent behavior. It excels in agentic-style execution, supporting a thinking-before-acting approach that improves performance across mainstream frameworks. Beyond its core coding capabilities, the model is optimized for UI generation, producing cleaner, modern webpages and accurately laid-out slides, making it a versatile tool for both technical and creative production workflows.

Building upon the lineage of the GLM-4.6, this iteration delivers substantial gains in multilingual coding, terminal-based tasks, and complex reasoning as evidenced by significant improvements on benchmarks like SWE-bench and Humanity’s Last Exam. The model demonstrates enhanced proficiency in tool-using scenarios and web browsing, ensuring reliable interaction with external systems. By unifying reasoning, coding, and agentic capabilities, it provides a coherent system that maintains performance across diverse scenarios, including chat and role-play, positioning it as a reliable engine for developers seeking to automate multi-step tasks in real-world production settings.