Qwen/Qwen3-Embedding-8B
LoRA-style vs Qwen/Qwen3-8B-Base. 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 36 of 36 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/145, MLP 108/108, Attention 144/144
32 derivatives checked
- Octen/Octen-Embedding-8B: Full fine-tune
- alexliap/Qwen3-Embedding-8B-NVFP4: Quantization
- drawais/Qwen3-Embedding-8B-AWQ-INT4: Quantization
- furiosa-ai/Qwen3-Embedding-8B: Identical copy
- alexliap/Qwen3-Embedding-8B-FP8-DYNAMIC: Quantization
- mlx-community/Qwen3-Embedding-8B-4bit-DWQ: Quantization
- Octen/Octen-Embedding-8B-INT8: Quantization
- mradermacher/Qwen3-Embedding-8B-i1-GGUF: Quantization
- mradermacher/Qwen3-Embedding-8B-GGUF: Same weights, converted
- huynguyendbs/Qwen3-Embedding-8B-4bit-MLX: Quantization
- LingoIITGN/qwen-indic-v1: Same weights, repacked
- QuaduxIT/Qwen3-Embedding-8B-W8A16: Quantization
- Mira190/Euler-Legal-Embedding-V1: LoRA-style
- boboliu/Qwen3-Embedding-8B-W4A16-G128: Quantization
- LostGentoo/Qwen3-Embedding-8B-AWQ: Quantization
- batiai/Qwen3-Embedding-8B-GGUF: Quantization
- cstr/qwen3-embed-8b-GGUF: Quantization
- jedisct1/Qwen3-Embedding-8B-mlx: Quantization
- emee-ai/Qwen3-Embedding-8B-4bit-DWQ: Quantization
- ThakiCloud/SKILLRET-Embedding-8B: Full fine-tune
- zenlm/zen-embedding-8B-GGUF: Quantization
- TiGa-RCE/Qwen3-Embedding-8B-MLX-BF16: Same weights, repacked
- LingoIITGN/FlowVec-v1: LoRA-style
- AutomatosX/AX-Qwen3-Embedding-8B-MLX-AXQ-MXFP8: Quantization
- sionic-ai/comsat-embed-ko-8b-preview: LoRA-style
- uw-math-ai/MathLeap-Qwen-8B: Full fine-tune
- sionic-ai/comsat-embed-ja-8b-preview: LoRA-style
- farbodtavakkoli/OTel-Embedding-8B: Full fine-tune
- enacimie/Qwen3-Embedding-8B-Q4_K_M-GGUF: Quantization
- yasinyaman/toolrank-emb-8b: LoRA-style
12 other repos hold byte-for-byte identical weight files.