mradermacher/Qwen3.5-4B-Claude-4.6-Opus-Reasoning-Distill-heretic-v3-GGUF
Same weights, converted vs meangrinch/Qwen3.5-4B-Claude-4.6-Opus-Reasoning-Distill-heretic-v3. Same weights stored in another dtype (BF16 → F16/F32), plus 24 new tensors.
Tensors changed: Embeddings 0/1, Norms 0/81, MLP 0/96, Attention 0/32