OpenGVLab/InternVL3-9B

Full fine-tune vs OpenGVLab/InternVL3-9B-Instruct. Fine-tune: 94% of the weights changed, across 48 of 48 layers.

Tensors changed: Attention 192/192, Norms 126/193, MLP 240/246, Embeddings 4/5, other 1/49

1 derivatives checked

• dmis-lab/InternVL3-9B-CVRR: Different architecture

1 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse