HEART77/TIR-Qwen3.5-4B-Thinking-v2

Full fine-tune vs Qwen/Qwen3.5-4B. Fine-tune: 93% of the weights changed, across 32 of 32 layers.

Tensors changed: Embeddings 1/4, MLP 96/196, Attention 187/320, Norms 68/203

1 derivatives checked

• mradermacher/TIR-Qwen3.5-4B-Thinking-v2-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse