Instructions to use d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp with MLX:

# Make sure mlx-vlm is installed
# pip install --upgrade mlx-vlm

from mlx_vlm import load, generate
from mlx_vlm.prompt_utils import apply_chat_template
from mlx_vlm.utils import load_config

# Load the model
model, processor = load("d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp")
config = load_config("d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp")

# Prepare input
image = ["http://images.cocodataset.org/val2017/000000039769.jpg"]
prompt = "Describe this image."

# Apply chat template
formatted_prompt = apply_chat_template(
    processor, config, prompt, num_images=1
)

# Generate output
output = generate(model, processor, formatted_prompt, image)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps Settings
LM Studio

How to use d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp"

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent

How to use d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp with Hermes Agent:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp"

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp"

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp" \
  --custom-provider-id mlx-lm \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Qwen3.8-Flash-Next-oQ4e-mtp

This model is an MLX quantization of Qwen/Qwen3.8-Flash-Next, made with oQe (oMLX v0.7.0) imatrix-weighted mixed-precision quantization.

See also the standard oQ version: d9beuD/Qwen3.8-Flash-Next-oQ4-mtp.

Quantization details

Base model: Qwen/Qwen3.8-Flash-Next
Model type: qwen4_exp
oQ level: 4
Default bits: 4
Group size: 64 (default; some modules use 32 or 128)
Effective size: ~4.72 bits per weight, 106 GB
dtype: bfloat16 (non-quantized weights, scales and biases)
MTP: multi-token prediction head preserved (mtp_num_hidden_layers: 1)
Vision encoder: included
N-gram embedding table: included
Format: MLX safetensors

Share of quantized parameters by bit width:

Bits	Share
8-bit	0.9%
6-bit	0.3%
5-bit	0.8%
4-bit	98.0%

Importance matrix (oQe)

Calibration set: oqe_code_multilingual, 1,024 samples x 512 tokens (adaptive: 128 to 1,024 samples, 8 rounds)
Collection: streamed one decoder layer at a time from the bf16 checkpoint (no proxy model)
Expert coverage: 75,240 / 75,264 routed experts activated (99.97%); the 24 experts that received no calibration tokens use standard oQ quantization
Full details in oq_imatrix_report.json

Sensitivity model

oQ's layer-sensitivity map was measured on Jundot/Qwen3.8-Flash-Next-oQ4e-mtp (128 samples x 256 tokens, code_multilingual calibration set) instead of the full bf16 checkpoint, which does not fit in memory on a 128 GB Mac.

Requirements

About 106 GB of weights: this model needs a Mac with more unified memory than that (e.g. 128 GB).

License

This model inherits the Qwen Community License 1.0 from the base model Qwen/Qwen3.8-Flash-Next.

Downloads last month: 164

Safetensors

Model size

180B params

Tensor type

U32

BF16

I64

MLX

Hardware compatibility

4-bit

Model tree for d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp

Base model

Qwen/Qwen3.8-Flash-Next

Quantized

(541)

this model

Collection including d9beuD/Qwen3.8-Flash-Next-oQ4e-mtp

Qwen3.8-Flash-Next oQe

Collection

9 items • Updated 26 days ago