Qwen3.8-27B-FP8

Qwen3.8-27B-FP8

<p> <a href="https://huggingface.co/Qwen/Qwen3.8-27B"><img src="https://img.shields.io/badge/Base_Model-Qwen3.8--27B-blue" alt="Base Model"></a> <a href="https://huggingface.co/Qwen/Qwen3.8-27B/blob/main/LICENSE"><img src="https://img.shields.io/badge/License-Apache_2.0-green" alt="License"></a> <img src="https://img.shields.io/badge/Quantization-FP8_Dynamic-orange" alt="FP8"> </p>

FP8 dynamically quantized version of Qwen/Qwen3.8-27B in safetensors format for efficient inference with reduced VRAM usage.

Model Details

Overview

This model applies FP8 dynamic quantization to all linear layers, reducing model size by approximately 50% while maintaining near-original accuracy. FP8 quantization provides the best quality-to-size ratio among quantization methods.

Usage

vLLM

vllm serve douyamv/Qwen3.8-27B-FP8 \
    --tensor-parallel-size 1 \
    --max-model-len 32768 \
    --trust-remote-code

SGLang

python -m sglang.launch_server \
    --model-path douyamv/Qwen3.8-27B-FP8 \
    --tp 1 \
    --trust-remote-code

Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "douyamv/Qwen3.8-27B-FP8",
    device_map="auto",
    trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained("douyamv/Qwen3.8-27B-FP8")

messages = [{"role": "user", "content": "Hello, tell me about yourself"}]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer([text], return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Base Model Information

• Model: Qwen3.8-27B
• Parameters: 27.78B
• Architecture: Hybrid (Gated DeltaNet + Gated Attention)
• Context Length: 262,144 tokens (extensible to 1M+)
• License: Apache 2.0
• Capabilities: Text generation, coding, reasoning, vision understanding, tool calling

Quantization Method

Quantized using llm-compressor with FP8_DYNAMIC scheme applied to all Linear layers.

Credits

• Original model by Qwen Team
• FP8 quantization by douyamv

Related Models

• douyamv/Qwen3.8-27B-GGUF — GGUF quantizations (Q2_K to Q8_0)
• douyamv/Qwen3.8-27B-abliterated — Abliterated (uncensored) version
• douyamv/Qwen3.8-27B-abliterated-GGUF — Abliterated GGUF quantizations