Qwen3.8-Flash-Next-CQ-4bit

Qwen3.8-Flash-Next · Coded Quantization (CQ) 4-bit

Note on Hugging Face's model-size display: Hugging Face may report a smaller stored parameter count for this repository because CQ stores model weights in a packed 4-bit representation. The displayed stored tensor count therefore does not directly represent the parameter count of the original model architecture.

Qwen3.8-Flash-Next quantized to 4-bit with Coded Quantization (CQ).

ThousandML: https://thousandml.com

CQ is a calibration-free 4-bit quantization method designed to preserve the behavior of the original model without requiring a calibration dataset. Because CQ does not optimize the quantization around a fixed calibration corpus, it is designed to generalize beyond calibration-specific data and workloads.

This repository contains Qwen3.8-Flash-Next quantized with Coded Quantization (CQ) 4-bit.

Why CQ?

Calibration-free. CQ does not require a representative calibration dataset before quantization.

This avoids tying the quantization procedure to a particular calibration corpus and is intended to provide better generalization beyond the data that would otherwise have been used for calibration.

4-bit inference. Model weights are quantized to 4-bit for substantially lower weight memory requirements than BF16/FP16 deployment.

Native vLLM integration. CQ models run through the CQ vLLM quantization plugin and custom CUDA kernels included in the Thousandml CQ Docker image.

Model

Run with Docker

The recommended way to run this model is with the dedicated Qwen3.8-Flash-Next CQ Docker image.

This image contains the compatible vLLM installation, CQ quantization plugin, CQ CUDA kernels, and runtime dependencies required by Qwen3.8-Flash-Next.

1. Make sure NVIDIA Docker support works

docker run --rm --gpus all \
  nvidia/cuda:13.0.2-base-ubuntu22.04 \
  nvidia-smi

Your GPU should appear in the output.

2. Start Qwen3.8-Flash-Next-CQ-4bit

docker run --rm --gpus all \
  -p 8000:8000 \
  docker.io/thousandml/cq:4bit-qwen38-flash-next-v1 \
  vllm serve thousandml/Qwen3.8-Flash-Next-CQ-4bit \
  --quantization cq

The model is then available through vLLM's OpenAI-compatible API on port 8000.

WSL2

If you are running Docker through WSL2 and vLLM reports that UVA is unavailable, enable vLLM's WSL2 pinned-memory support:

docker run --rm --gpus all \
  -e VLLM_WSL2_ENABLE_PIN_MEMORY=1 \
  -p 8000:8000 \
  docker.io/thousandml/cq:4bit-qwen38-flash-next-v1 \
  vllm serve thousandml/Qwen3.8-Flash-Next-CQ-4bit \
  --quantization cq

Send a request

Once the server is ready:

curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "thousandml/Qwen3.8-Flash-Next-CQ-4bit",
    "messages": [
      {
        "role": "user",
        "content": "Explain quantization in one paragraph."
      }
    ]
  }'

Reasoning and tool calling

Qwen3.8-Flash-Next can be served with vLLM's reasoning and automatic tool-choice support.

docker run --rm --gpus all \
  -p 8000:8000 \
  docker.io/thousandml/cq:4bit-qwen38-flash-next-v1 \
  vllm serve thousandml/Qwen3.8-Flash-Next-CQ-4bit \
  --quantization cq \
  --enable-auto-tool-choice \
  --tool-call-parser qwen3_xml \
  --reasoning-parser qwen3

Hugging Face authentication

For higher Hugging Face Hub rate limits, pass your Hugging Face token into the container:

docker run --rm --gpus all \
  -e HF_TOKEN="$HF_TOKEN" \
  -p 8000:8000 \
  docker.io/thousandml/cq:4bit-qwen38-flash-next-v1 \
  vllm serve thousandml/Qwen3.8-Flash-Next-CQ-4bit \
  --quantization cq

On WSL2, combine it with:

-e VLLM_WSL2_ENABLE_PIN_MEMORY=1

Full recommended command

For deployments using reasoning, automatic tool calling, and the full supported context length:

docker run --rm --gpus all \
  -e HF_TOKEN="$HF_TOKEN" \
  -p 8000:8000 \
  docker.io/thousandml/cq:4bit-qwen38-flash-next-v1 \
  vllm serve thousandml/Qwen3.8-Flash-Next-CQ-4bit \
  --quantization cq \
  --max-model-len 262144 \
  --enable-auto-tool-choice \
  --tool-call-parser qwen3_xml \
  --reasoning-parser qwen3

For WSL2, add:

-e VLLM_WSL2_ENABLE_PIN_MEMORY=1

About Coded Quantization (CQ)

Learn more at https://thousandml.com.

Coded Quantization (CQ) is a calibration-free quantization approach for efficient LLM inference.

Traditional post-training quantization methods can depend on a calibration dataset to estimate quantization parameters. The resulting quantization can therefore be influenced by the distribution and composition of that calibration data.

CQ removes that calibration-data requirement. The goal is to retain strong fidelity to the original model while allowing the quantized representation to generalize beyond any particular calibration corpus.

Base model

This checkpoint is derived from Qwen/Qwen3.8-Flash-Next.

Please refer to the original Qwen model card for architecture details, capabilities, usage guidance, limitations, and upstream licensing information.

License

The base model is released under the applicable upstream Qwen license terms. See the repository license and the original Qwen model card for details.