Qwen3.8-27B-NF4 (Experimental)

Experimental repository: NF4 quantization of Qwen/Qwen3.8-27B produced with AGIWSNeuralQuant.

• Base model: Qwen/Qwen3.8-27B — 27B dense vision-language model (image-text-to-text), Apache 2.0
• Quantization: NF4 (NormalFloat 4-bit, per-group, double quantization), weights-only; norms/embeddings/vision kept in higher precision
• Tooling: AGIWSNeuralQuant (pure PyTorch, no bitsandbytes)

This repository is an experimental workspace for the quantization pipeline. Weights will be uploaded as the conversion progresses. Do not rely on it for production.