license: apache-2.0 base_model: orcarouter/Qwen3.8-27B-Uncensored-FP8 pipeline_tag: image-text-to-text tags:

• transformers
• safetensors
• qwen3_5
• fp8
• block-fp8
• vllm
• nvidia

Qwen3.8-27B-Uncensored-FP8 (NVIDIA / FP8)

Unofficial mirror of orcarouter/Qwen3.8-27B-Uncensored-FP8 in the original FP8 safetensors format, intended for NVIDIA GPUs (vLLM / transformers with FP8 support).

Format

• transformers checkpoint, qwen3_5 architecture (Qwen3_5ForConditionalGeneration)
• 7 safetensors shards, FP8 (block-FP8) quantized weights
• Multimodal (image-text-to-text)

Usage (NVIDIA / vLLM)

# vLLM (FP8 native)
vllm serve id-2/Qwen3.8-27B-Uncensored-FP8 --quantization fp8

# transformers
from transformers import AutoModelForCausalLM, AutoProcessor
model = AutoModelForCausalLM.from_pretrained("id-2/Qwen3.8-27B-Uncensored-FP8")

Note: MLX format (Apple Silicon) is NOT provided here — this is the NVIDIA/FP8 build.