[trollbridge]

GLM-5.2 just keeps blowing me away. I pretty much use it in the mode I used to use Opus for where I have it drive a lot of the reasoning with DeepSeek or other models for the smaller subtasks/subagents.