kurakurai/Luth-2-2B

Partial fine-tune vs Qwen/Qwen3.5-2B. Partial change: 85% of the weights changed (attention, embeddings, mlp, norms), in 24 of 24 layers.

Tensors changed: Attention 155/264, Norms 77/177, Embeddings 1/4, MLP 72/172

2 derivatives checked

• mradermacher/Luth-2-2B-i1-GGUF: Quantization
• mradermacher/Luth-2-2B-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse