SyzygyResearch/Mach-2-Additive-Medium
Partial fine-tune (fewer layers) vs Qwen/Qwen3.8-Flash-Next. SyzygyResearch/Mach-2-Additive-Medium matches part of Qwen/Qwen3.8-Flash-Next (560 of its 1658 tensors; the rest of Qwen/Qwen3.8-Flash-Next has no counterpart). On the shared part: Partial change: 82% of the weights changed (mlp, norms), in 48 of 48 layers.
Tensors changed: Norms 160/184, Attention 0/228, MLP 96/144, Embeddings 0/3, other 0/1
1 derivatives checked
- SyzygyResearch/Mach-2-Additive-Medium-GGUF: Not comparable yet