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

Libraries

How to use Jundot/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("Jundot/Qwen3.8-Flash-Next-oQ4e-mtp")
config = load_config("Jundot/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 Jundot/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 "Jundot/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": "Jundot/Qwen3.8-Flash-Next-oQ4e-mtp"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent

How to use Jundot/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 "Jundot/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 Jundot/Qwen3.8-Flash-Next-oQ4e-mtp

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use Jundot/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 "Jundot/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 "Jundot/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

Updated 2026-10-11: Re-quantized with Improved oQe Quantization. Please re-download if you have an earlier copy.

This model was quantized using oQ (oMLX v0.7.1) mixed-precision quantization with Improved oQe Quantization (jundot/omlx#4385).

A mixed-precision MLX checkpoint of Qwen/Qwen3.8-Flash-Next, an experimental preview of the architecture behind Qwen4: a mixture-of-experts vision-language model with 125B parameters and 6B activated, plus a 51B n-gram embedding and a 4B MTP module, that pairs Gated DeltaNet with Qwen Sparse Attention (QSA) and uses Gated Residual streams. It includes the language model, the n-gram embedding, the vision encoder and the multi-token prediction (MTP) head.

The effective average is 4.724 bits/weight for the complete checkpoint.

KL Divergence from the Original Model

Mean next-token KL divergence from the original BF16 model on the oMLX benchmark corpora (lower is better). Each corpus is scored on 16 windows of 2,048 tokens spread across the text (32,768 tokens).

Corpus	Previous oQ4e (2026-08-27)	Updated oQ4e	Change
Code (Python)	0.0877	0.0776	-11.6%
Novel (English)	0.1909	0.1713	-10.2%

Quantization and Bit Distribution

oQ4e allocates bits by measured layer sensitivity, then rounds each group with an importance matrix (1024 samples x 512 tokens) and a weighted least-squares refit of its scale and bias.

Storage format	Logical weights	Share of total	Tensor storage	Effective bits/weight
Affine 4-bit, group 32	51.200B	28.445%	29.802 GiB	5.00
Affine 4-bit, group 64	124.724B	69.291%	65.339 GiB	4.50
Affine 5-bit, group 128	0.981B	0.545%	0.600 GiB	5.25
Affine 5-bit, group 64	0.302B	0.168%	0.193 GiB	5.50
Affine 6-bit, group 128	0.003B	0.001%	0.002 GiB	6.25
Affine 6-bit, group 64	0.629B	0.349%	0.476 GiB	6.50
Affine 8-bit, group 128	0.241B	0.134%	0.231 GiB	8.25
Affine 8-bit, group 64	1.405B	0.780%	1.390 GiB	8.50
I64	0.000B	0.000%	0.000 GiB	64.00
BF16	0.516B	0.287%	0.961 GiB	16.00
Total	180.000B	100%	98.994 GiB	4.724

Storage Breakdown

Component	Logical weights	Storage
Language backbone	125.744B	71.913 GB / 66.974 GiB
N-gram embedding	51.200B	32.000 GB / 29.802 GiB
Vision encoder	0.449B	0.898 GB / 0.836 GiB
MTP head	2.607B	1.483 GB / 1.381 GiB
Total	180.000B	106.294 GB / 98.994 GiB

License

Qwen Community License 1.0, following the source model. See LICENSE.

Downloads last month: 28,756

Safetensors

Model size

180B params

Tensor type

U32

BF16

I64

MLX

Hardware compatibility

4-bit

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

Base model

Qwen/Qwen3.8-Flash-Next

Quantized

(483)

this model

Quantizations

1 model