GLM 5.2: What Makes it So Special?

Short summary

GLM-5.2 is a 744B parameter open-weight MoE model with a 1M token context window and sparse attention optimization that reduces compute by 2.9× at full context. Its 40B active parameters enable strong coding performance (74.4% on Frontier SWE) with efficient inference and multi-token prediction. MIT-licensed with self-hosting capability, it's cost-competitive with closed US models.

• •744B open-weight MoE with 1M context, sparse attention (2.9× fewer ops at full context)
• •74.4% Frontier SWE performance, 20% faster inference via multi-token prediction
• •MIT-licensed, self-hostable, pricing competitive with closed models

Generated with AI, which can make mistakes.