karakuri-ai/karakuri-vl-2-8b-thinking-2603

Partial fine-tune vs Qwen/Qwen3-VL-8B-Thinking. Partial change: 86% of the weights changed (attention, lm_head, mlp, norms), in 36 of 36 layers.

Tensors changed: Embeddings 0/4, Norms 140/261, MLP 108/232, Attention 144/252, lm_head 1/1

3 derivatives checked

• mradermacher/karakuri-vl-2-8b-thinking-2603-i1-GGUF: Quantization (fewer layers)
• mradermacher/karakuri-vl-2-8b-thinking-2603-GGUF: Same weights, converted (fewer layers)
• mmnga-o/karakuri-vl-2-8b-thinking-2603-gguf: Quantization (fewer layers)

On the Hub · Downloads on Model Pulse