neopolita/Qwen3.8-Flash-Next-99B-A5B-Niwaki-3bit-mlx

Partial fine-tune vs mlx-community/Qwen3.8-Flash-Next-4bit. Partial change: 5% of the weights changed (mlp), in 36 of 36 layers.

Tensors changed: Embeddings 0/6, Attention 0/944, Norms 0/138, MLP 263/864, other 0/7

1 derivatives checked

• neopolita/Qwen3.8-Flash-Next-99B-A5B-Niwaki-3bit-gguf: Quantization

On the Hub · Downloads on Model Pulse