WeiboAI/VibeThinker-1.5B
Full fine-tune vs Qwen/Qwen2.5-Math-1.5B. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 1/1, Norms 57/57, MLP 84/84, Attention 196/196
13 derivatives checked
- prithivMLmods/VibeThinker-1.5B-f32-GGUF: Same weights, converted
- constructai/VibeThinker-1.5B-GGUF: Same weights, converted
- mradermacher/VibeThinker-1.5B-i1-GGUF: Quantization
- mradermacher/VibeThinker-1.5B-GGUF: Same weights, converted
- DevQuasar/WeiboAI.VibeThinker-1.5B-GGUF: Same weights, converted
- MaziyarPanahi/VibeThinker-1.5B-GGUF: Same weights, converted
- squ11z1/Chronos-1.5B: Different architecture
- mlx-community/VibeThinker-1.5B-mlx-4bit: Quantization
- sirunchained/VibeThinker-1.5B-gguf: Same weights, converted
- raithdemis/Qwen2.5-1.5B-VibeThinker-heretic-uncensored-abliterated: Abliterated
- DavidAU/Qwen2.5-1.5B-VibeThinker-heretic-uncensored-abliterated: Abliterated
- OpceanAI/Yuuki-RxG-nano: LoRA-style
- prometheus04/vibethinker-1.5b-rstar-coder: LoRA-style
1 other repos hold byte-for-byte identical weight files.