cerebras/MiniMax-M2-REAP-139B-A10B
Partial fine-tune (fewer layers) vs MiniMaxAI/MiniMax-M2. cerebras/MiniMax-M2-REAP-139B-A10B matches part of MiniMaxAI/MiniMax-M2 (58035 of its 96103 tensors; the rest of MiniMaxAI/MiniMax-M2 has no counterpart). On the shared part: Partial change: 37% of the weights changed (mlp), in 62 of 62 layers.
Tensors changed: MLP 1263/1267, Attention 0/496, Embeddings 0/1, Norms 0/249, lm_head 0/1
3 derivatives checked
- bartowski/cerebras_MiniMax-M2-REAP-139B-A10B-GGUF: Quantization
- DevQuasar/cerebras.MiniMax-M2-REAP-139B-A10B-GGUF: Quantization
- mradermacher/MiniMax-M2-REAP-139B-A10B-i1-GGUF: Quantization