The GLM-4.6 model now officially supports the thinking feature in Claude Code, bringing more powerful reasoning capabilities to AI applications. According to official documentation, the model’s /chat/completions endpoint enables the thinking feature by default, with the model autonomously determining whether thinking is necessary. Researchers have successfully adapted the thinking display feature for CCR by splitting thinking parameters. For configuration, users need to use the latest version of CCR and set it up through the OpenAI-format Coding endpoint, with example configurations provided. Although the interleaved thinking feature is temporarily missing, this update significantly improves the GLM model’s performance in complex tasks, providing developers with more flexible AI tool options.
