ggml-org/Qwen2.5-VL-32B-Instruct-GGUF

Same weights, converted (fewer layers) vs Qwen/Qwen2.5-VL-32B-Instruct. ggml-org/Qwen2.5-VL-32B-Instruct-GGUF matches part of Qwen/Qwen2.5-VL-32B-Instruct (771 of its 1161 tensors; the rest of Qwen/Qwen2.5-VL-32B-Instruct has no counterpart). On the shared part: Same weights stored in another dtype (BF16/F32 → F16/F32).

Tensors changed: Embeddings 0/1, Norms 0/129, MLP 0/192, Attention 0/448, lm_head 0/1

On the Hub · Downloads on Model Pulse