Z.ai Releases 753B GLM-5.2 Open Weights With a 1M-Token Context
Z.ai made the GLM-5.2 weights available on Hugging Face under the MIT license in BF16 and F32 safetensors formats. The MoE model has 753B total and 40B active parameters. IndexShare reuses an indexer across four sparse attention layers, reducing FLOPs per token by a factor of 2.9 at 1M context. Coverage described strong results in long-running coding and agent tasks at roughly one-sixth the cost of GPT-5.5.
Visit original source