njmason/caveman-qwen3.6-GGUF
Adapter vs njmason/caveman-qwen3.6. njmason/caveman-qwen3.6 is a LoRA adapter, not full weights: rank 32, on k_proj, o_proj, q_proj, v_proj in 10 layers (40 adapted matrices). It modifies njmason/caveman-qwen3.6-GGUF at load time.