Qwen/Qwen3.5-2B
Partial fine-tune vs Qwen/Qwen3.5-2B-Base. Partial change: 87% of the weights changed (attention, embeddings, mlp, norms, other), in 24 of 24 layers.
Tensors changed: Attention 146/268, Norms 87/183, Embeddings 3/5, MLP 79/175, other 1/1
262 derivatives checked
- unsloth/Qwen3.5-2B-GGUF: Same weights, converted
- unsloth/Qwen3.5-2B: Identical copy
- bartowski/Qwen_Qwen3.5-2B-GGUF: Same weights, converted
- lmstudio-community/Qwen3.5-2B-GGUF: Quantization
- Alibaba-DAMO-Academy/RynnBrain1.1-2B: Full fine-tune
- unsloth/Qwen3.5-2B-MTP-GGUF: Same weights, converted
- tencent/WeMM-Embedding-2B: Full fine-tune
- OpenMed/Qwen3.5-2B-MedVL: LoRA-style
- mlx-community/Qwen3.5-2B-MLX-4bit: Quantization
- atakhadivi/Qwen3.8-2B-Uncensored-GGUF: Full fine-tune
- NateRunsA-LIST/alist-food-text-qwen35: Not comparable yet
- cyankiwi/Qwen3.5-2B-AWQ-4bit: Quantization
- lmstudio-community/Qwen3.5-2B-MLX-4bit: Quantization
- empero-ai/Qwen3.8-2B-Distill: Partial fine-tune
- techwithsergiu/Qwen3.5-text-2B: Same weights, repacked (fewer layers)
- saidutta69/Qwen3.5-2B-heretic: Abliterated
- topk-io/topk-embed-v1-small: LoRA-style
- lmstudio-community/Qwen3.5-2B-MLX-8bit: Quantization
- diodel/Qwen3.5-2B-Q4_K_M-GGUF: Quantization
- tuandunghcmut/Qwen3.5-2B-text-only: Same weights, converted (fewer layers)
- mlx-community/Qwen3.5-2B-MLX-8bit: Quantization
- trubisky/Qwen3.5-2B-RCOL: Quantization
- huihui-ai/Huihui-Qwen3.5-2B-abliterated: Abliterated
- AxionML/Qwen3.5-2B-NVFP4: Quantization
- Freesolo-Co/Qwen3.5-2B-FP8: Quantization
- internlm/Intern-Decision-2B: Partial fine-tune
- UT-IFML/dQwen3.5-2B-Base: Full fine-tune (fewer layers)
- AaryanK/Qwen3.5-2B-GGUF: Same weights, converted
- trl-lab/qwen3.5-2b-grpo-sqale: Partial fine-tune
- ML-Intern-lab/Qwen-Image-2.1-PE-T2I-Pocket-2B: Full fine-tune (fewer layers)
35 other repos hold byte-for-byte identical weight files.