0xSero/Qwen3.5-28B

Same weights, converted vs Qwen/Qwen3.5-35B-A3B. Same weights stored in another dtype (BF16/F32 → BF16), plus 24600 new tensors.

Tensors changed: lm_head 0/1, Embeddings 0/4, Attention 0/388, MLP 0/272, Norms 0/241

3 derivatives checked

• mradermacher/Qwen-3.5-28B-A3B-REAP-i1-GGUF: Quantization
• barozp/Qwen-3.5-28B-A3B-REAP-GGUF: Same weights, converted
• mradermacher/Qwen-3.5-28B-A3B-REAP-GGUF: Quantization

On the Hub · Downloads on Model Pulse