logic65/Whittle-Qwen-3.8-35B-A3B
Full fine-tune vs logic65/Whittle-Next-27B-A3B. Fine-tune: 95% of the weights changed, across 40 of 40 layers.
Tensors changed: lm_head 0/1, Embeddings 3/5, Attention 250/410, MLP 399/400, Norms 84/154, other 5/5
5 derivatives checked
- logic65/Whittle-Qwen-3.8-35B-A3B-GGUF: Not comparable yet
- mradermacher/Whittle-Qwen-3.8-35B-A3B-i1-GGUF: Not comparable yet
- el4/Whittle-Qwen-3.8-35B-A3B-ONYX-GGUF: Not comparable yet
- mradermacher/Whittle-Qwen-3.8-35B-A3B-GGUF: Not comparable yet
- logic65/Whittle-Reflex-25B-A3B: Partial fine-tune