KennethFal/16bit-pixel-lora-minimax-h3
Adapter vs MiniMaxAI/MiniMax-H3. KennethFal/16bit-pixel-lora-minimax-h3 is a LoRA adapter, not full weights: rank 32, on out_proj, qkv_proj in 50 layers (104 adapted matrices). It modifies MiniMaxAI/MiniMax-H3 at load time.
1 other repos hold byte-for-byte identical weight files.