●builderGLM-5.2 offers GPT-5.4-comparable reasoning at $0.25 per ARC-AGI-2 task, worth evaluating for cost-sensitive reasoning workloads.
●researcherARC-AGI-2 scores now provide a cleaner signal on reasoning capability gaps between closed and open-weight models, with GLM-5.2 as a concrete open-weights reference point.
●founderWorth watching because the 8-12 month closed-to-open lag means open-weight alternatives to frontier reasoning models are arriving faster than prior cycles.