Skywork/Skywork-R1V3-38B

Full fine-tune vs OpenGVLab/InternVL3-38B. Fine-tune: 97% of the weights changed, across 64 of 64 layers.

Tensors changed: Embeddings 3/5, Norms 265/309, Attention 583/583, other 90/90, MLP 371/378, lm_head 1/1

2 derivatives checked

• bartowski/Skywork_Skywork-R1V3-38B-GGUF: Same weights, converted (fewer layers)
• lmstudio-community/Skywork-R1V3-38B-GGUF: Quantization (fewer layers)

On the Hub · Downloads on Model Pulse