zlab-princeton/Vero-Qwen3I-8B

Full fine-tune vs Qwen/Qwen3-VL-8B-Instruct. Fine-tune: 93% of the weights changed, across 36 of 36 layers.

Tensors changed: Embeddings 3/4, Norms 188/261, MLP 224/232, Attention 252/252, lm_head 1/1

2 derivatives checked

• mradermacher/Vero-Qwen3I-8B-i1-GGUF: Quantization (fewer layers)
• mradermacher/Vero-Qwen3I-8B-GGUF: Same weights, converted (fewer layers)

On the Hub · Downloads on Model Pulse