mradermacher/GLM-5.2-0.8B-A0.8B-GGUF

Same weights, converted vs inference-optimization/GLM-5.2-0.8B-A0.8B. Same weights stored in another dtype (F32 → F16/F32), plus 28 new tensors.

Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/31, MLP 0/22, Attention 0/33

On the Hub · Downloads on Model Pulse