unsloth/Qwen3.5-2B-Base

Same weights, converted vs Qwen/Qwen3.5-2B-Base. Same weights stored in another dtype (BF16/F32 → BF16).

Tensors changed: Attention 0/264, Norms 0/177, Embeddings 0/4, MLP 0/172

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

On the Hub · Downloads on Model Pulse