Qwen/Qwen3-4B
Full fine-tune vs Qwen/Qwen3-4B-Base. Fine-tune: 100% of the weights changed, across 36 of 36 layers.
Tensors changed: Embeddings 1/1, Norms 145/145, MLP 108/108, Attention 144/144
421 derivatives checked
- zeroentropy/zerank-2-reranker: LoRA-style
- Qwen/Qwen3-4B-GGUF: Quantization
- unsloth/Qwen3-4B-GGUF: Same weights, converted
- AnkitAI/Parable-Qwen3-4B-Claude-Fable-5-GGUF: LoRA-style
- bartowski/Qwen_Qwen3-4B-GGUF: Same weights, converted
- Qwen/Qwen3-4B-AWQ: Quantization
- MaziyarPanahi/Qwen3-4B-GGUF: Same weights, converted
- unsloth/Qwen3-4B: Same weights, repacked
- Qwen/Qwen3-4B-FP8: Quantization
- Qwen/Qwen3Guard-Gen-4B: Full fine-tune
- arianraje/qwen3-4b-mamba3-hybrid-stage2b-kd-bias0: Full fine-tune
- unsloth/Qwen3-4B-unsloth-bnb-4bit: Quantization
- zeroentropy/zerank-1-small-reranker: Different architecture
- mlx-community/Qwen3-4B-4bit: Quantization
- lmstudio-community/Qwen3-4B-MLX-4bit: Quantization
- lmstudio-community/Qwen3-4B-MLX-8bit: Quantization
- electricpipelines/Qwen3-4B-GGUF: Full fine-tune
- YannQi/R-4B: Full fine-tune (fewer layers)
- lmstudio-community/Qwen3-4B-GGUF: Quantization
- ggml-org/Qwen3-4B-GGUF: Same weights, converted
- unsloth/Qwen3-4B-bnb-4bit: Quantization
- DreamFast/qwen3-4b-heretic: Abliterated
- mlx-community/Qwen3-4B-8bit: Quantization
- zeroentropy/zembed-1-embedding: LoRA-style
- enacimie/Qwen3-4B-Q4_K_M-GGUF: Quantization
- gohumanize/gohumanize-open-humanizer: Full fine-tune
- RedHatAI/Qwen3-4B-quantized.w4a16: Quantization
- geoffmunn/Qwen3-4B-f16: Not comparable yet
- unsloth/Qwen3-4B-128K-GGUF: Same weights, converted
- NobodyWho/Qwen_Qwen3-4B-GGUF: Quantization
41 other repos hold byte-for-byte identical weight files.