ukisai/Swift-Qwen3.8-27b
LoRA-style vs Qwen/Qwen3.8-27B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 63 of 64 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/325, Attention 64/560, MLP 192/307, Embeddings 0/5, lm_head 0/1, other 0/1
56 derivatives checked
- ukisai/Swift-Qwen3.8-27B-GGUF: Same weights, converted
- bartowski/ukisai_Swift-Qwen3.8-27b-GGUF: Same weights, converted
- ukisai/Swift-Qwen3.8-27B-NVFP4: Quantization
- ticeclock/Swift-Qwen3.8-27B-RCO-GGUF: Quantization
- Yanun/Swift-Qwen3.8-27b-oQ4e-fp16-mtp: Quantization
- vmarcelo/Swift-Qwen3.8-27B-MIX_GGUF: Quantization
- TheUnderscore/Swift-Qwen3.8-27b-W4A16-AWQ: Quantization
- d0xin/Swift-Qwen3.8-27B-FP8: Quantization
- cyankiwi/Swift-Qwen3.8-27B-AWQ-INT4: Quantization
- LuffyTheFox/Swift-Qwen3.8-27B-Genesis-GGUF: LoRA-style
- mradermacher/Swift-Qwen3.8-27b-i1-GGUF: Quantization
- mradermacher/Swift-Qwen3.8-27b-GGUF: Quantization
- Yanun/Swift-Qwen3.8-27b-oQ4e-mtp: Quantization
- arsis-dev/ukisai-Swift-Qwen3.8-27b-MTPLX: Quantization
- kataguru/Swift-Qwen3.8-27B-Finnish-GGUF: Quantization
- ajgazin/Swift-Qwen3.8-27B-Uncensored-MTP: Abliterated
- ukisai/Swift-Qwen3.8-27b-int4-AMD: Quantization
- jamesbrunet/Swift-Qwen3.8-27b-W4A16-AutoRound: Quantization
- Dragoy/Swift-Qwen3.8-27B-abliterated-NVFP4: Quantization
- andosen/Swift-Qwen3.8-27b-mlx-5Bit: Quantization
- SixVolts/Swift-Qwen3.8-27B-GGUF: Quantization
- TokenAI-zer/Swift-Qwen3.8-27b-oQ4-mtp: Quantization
- d0xin/Swift-Qwen3.8-27B-Uncensored-BF16: Abliterated
- Dragoy/Swift-Qwen3.8-27B-abliterated-NVFP4-DFlash2: Quantization
- andrewting/Swift-Qwen3.8-27B-Abliterated: Abliterated
- TokenAI-zer/Swift-Qwen3.8-27b-oQ6-mtp: Quantization
- ahmedelkilani01/Swift-Qwen3.8-27B-i1-IQ4_XS-Smaller-GGUF: Quantization
- majentik/Swift-Qwen3.8-27B-MLX-6bit: Quantization
- SWiesmann/ukisai-Swift-Qwen3.8-27b-Q6-FP16-MTPLX: Quantization
- bjonor/Swift-Qwen3.8-27B-GPTQ-Int4-sym-G128-MTP-BF16: Quantization
1 other repos hold byte-for-byte identical weight files.