ukuwzi/Qweble-Sol-4B-MLX

Partial fine-tune (fewer layers) vs Qwen/Qwen3.5-4B. ukuwzi/Qweble-Sol-4B-MLX matches part of Qwen/Qwen3.5-4B (402 of its 738 tensors; the rest of Qwen/Qwen3.5-4B has no counterpart). On the shared part: Partial change: 5% of the weights changed (attention, mlp, norms), in 32 of 32 layers.

Tensors changed: Embeddings 0/1, MLP 6/96, Attention 9/200, Norms 81/105

On the Hub · Downloads on Model Pulse