I'm seeing a lot of people dunk on GLM-5.2. "too slow" "too expensive" too verbose" Yes, this model has its flaws, but let's not act like this is not the most impressive LLM release we've seen all year *aside from Fable* GLM-5.2 represented a 2x increase in open-sourced intelligence capabilities overnight. This is not meant to be a general conversational model like Claude/GPT (hence why it's extremely verbose). It's meant for one thing: Running coding tasks inside a specific AI harness. For general tasks, you should be running other models/subagents (DeepSeek, Qwen, Kimi). The GLM pricing plan is also extremely reasonable, as it's billed per prompt per session. This is a huge leap forward for open-sourced intelligence and AI as a whole. I won't be convinced otherwise.(Miles Deutscher)