Qwen3.8-27B-NVFP4

NVFP4 quantization of Qwen/Qwen3.8-27B, produced with llm-compressor via oneshot calibration.

Vision encoder, gated-delta attention (conv1d / linear_attn), lm_head, embeddings, and norm layers are kept in original precision.

Downloads last month
21
Safetensors
Model size
19B params
Tensor type
BF16
·
U8
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for lokeshe09/Qwen3.8-27B-NVFP4

Base model

Qwen/Qwen3.8-27B
Quantized
(386)
this model