OpenMOSE/GLM-5.3-Flash-REAP-250B-A18B

Partial fine-tune vs zai-org/GLM-5.3-Flash. Partial change: 18% of the weights changed (mlp), in 42 of 46 layers.

Tensors changed: MLP 904/1501, lm_head 0/1, Embeddings 0/3, other 0/4, Norms 0/277, Attention 0/875

On the Hub · Downloads on Model Pulse