GLM-4.7 is an advanced AI model that serves as a coding partner with enhanced capabilities across multiple domains. It represents a significant step forward in artificial intelligence, particularly focused on coding and reasoning tasks.

GLM-4.7 brings clear gains in multilingual agentic coding and terminal-based tasks, achieving 73.8% on SWE-bench (+5.8% over GLM-4.6), 66.7% on SWE-bench Multilingual (+12.9%), and 41% on Terminal Bench 2.0 (+16.5%). The model supports thinking before acting with significant improvements on complex tasks in mainstream agent frameworks including Claude Code, Kilo Code, Cline, and Roo Code. It also delivers substantial improvements in tool using, mathematical reasoning, and produces cleaner, more modern webpages with better-looking slides featuring accurate layout and sizing.

The model features Preserved Thinking technology that maintains reasoning context across multi-turn agentic tasks. This allows the AI to maintain continuity and context throughout extended interactions, making it particularly effective for complex, multi-step problem-solving scenarios.

GLM-4.7 shows significant improvements in many scenarios including chat, creative writing, and role-play scenarios. It achieves 42.8% on the HLE benchmark (+12.4% over GLM-4.6) and demonstrates strong performance across 17 different benchmarks covering reasoning, coding, and agent capabilities.

This model is designed for developers and AI practitioners working on complex coding tasks, mathematical reasoning problems, and multi-turn agentic applications. It's compatible with various tools and frameworks and available through multiple platforms including web interfaces and API access.