Z.ai GLM 5.2 is a large-scale reasoning model that excels at coding and tool use across long-running tasks. It maintains engineering context and follows standards consistently through a full development workflow. Accessed via OpenRouter with a context window of up to 1M tokens. Well suited for long-horizon agent workflows, project-level software engineering, and complex multi-step automation. Use the reasoning-effort dropdown to tune thinking depth; pick high or xhigh for hard problems. For everyday chat, Seed or the local models cost less; for deep Chinese-language work, consider Qwen 3.5. Prompt example 1: Build a Python command-line to-do tool with add, complete, delete and filter-by-date commands, storing data in a JSON file, plus usage instructions. Prompt example 2: We have a single 2,000-line order processing file. Propose steps to refactor it into layers: module split, interface design, test coverage and migration order. Prompt example 3: Write version control branching and code review guidelines for our team: branch naming, commit message format, merge criteria and a review checklist. Prompt example 4: Write a Rust command-line tool that recursively scans a folder, finds files with duplicate content by hash and prints them in groups, with build and usage instructions. Prompt example 5: Design a RESTful API for a library lending system: endpoints, HTTP methods, request and response JSON samples, and error code conventions. Prompt example 6: After scaling our Node.js service to four instances, order IDs sometimes repeat; they come from an in-memory counter (let seq = 0). Explain why and give three fixes with trade-offs. Prompt example 7: Write a Go monitor that checks https://shop.example.com every five minutes, posts an alert to https://hooks.example.com/alert after three straight failures, and logs each check. Prompt example 8: Plan an upgrade from MySQL 5.7 to 8.0, including a compatibility checklist, a backup and rollback plan, and ways to minimize downtime. Prompt example 9: Write a Git pre-commit hook that runs format checks and unit tests before each commit, aborts the commit if either fails, and explain how to share it with the team. Prompt example 10: Our 40-page React site needs Chinese, English and Japanese language switching. Plan the i18n architecture, translation file layout and migration order, and show how to convert one component. Maximum context window (input and output combined): about 1,048,576 tokens. Maximum output per reply: about 65,500 tokens (model reasoning, if any, counts toward this limit). Maximum total upload size per request: 512MB (all files and message content combined). Output content formats: Text Pricing: Medium Continuous conversation: Not supported AI can make mistakes. Please verify important information.