Instructions to use lukaskremla/Qwen3.8-27B-MTP-bf16-MLX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lukaskremla/Qwen3.8-27B-MTP-bf16-MLX with MLX:

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

# Generate text with mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("lukaskremla/Qwen3.8-27B-MTP-bf16-MLX")

prompt = "Write a story about Einstein"
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True
)

text = generate(model, tokenizer, prompt=prompt, verbose=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings
LM Studio

How to use lukaskremla/Qwen3.8-27B-MTP-bf16-MLX with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "lukaskremla/Qwen3.8-27B-MTP-bf16-MLX"

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": "lukaskremla/Qwen3.8-27B-MTP-bf16-MLX"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

MLX LM

How to use lukaskremla/Qwen3.8-27B-MTP-bf16-MLX with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "lukaskremla/Qwen3.8-27B-MTP-bf16-MLX"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "lukaskremla/Qwen3.8-27B-MTP-bf16-MLX"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "lukaskremla/Qwen3.8-27B-MTP-bf16-MLX",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Hermes Agent

How to use lukaskremla/Qwen3.8-27B-MTP-bf16-MLX 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 "lukaskremla/Qwen3.8-27B-MTP-bf16-MLX"

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 lukaskremla/Qwen3.8-27B-MTP-bf16-MLX

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use lukaskremla/Qwen3.8-27B-MTP-bf16-MLX with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "lukaskremla/Qwen3.8-27B-MTP-bf16-MLX"

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 "lukaskremla/Qwen3.8-27B-MTP-bf16-MLX" \
  --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"

lukaskremla/Qwen3.8-27B-MTP-bf16-MLX

This repository contains the native Multi-Token Prediction (MTP) tensors extracted from Qwen/Qwen3.8-27B and converted into the standalone MLX drafter format expected by mlx-vlm.

This is not a standalone language model. It must be loaded as the draft model alongside a compatible Qwen3.8-27B target model.

The same BF16 drafter can be used with the 2-bit, 3-bit, 4-bit, 5-bit, 6-bit and 8-bit MLX target models. Quantizations don't need to match.

You can download the vision capable or text-only Qwen 3.8 27B weights in a variety of MLX quantizations from this collection Qwen 3.8 27B MLX-Quants (Vision, Text-Only & MTP).

Hugging Face might render incorrect parameter counts for this model, it is a common display bug for MLX quants.

Credits

Converted to MLX format from Qwen/Qwen3.8-27B using mlx-vlm version 0.6.13.

Downloads last month: 146

Safetensors

Model size

0.4B params

Tensor type

BF16

MLX

Hardware compatibility

Quantized

Collection including lukaskremla/Qwen3.8-27B-MTP-bf16-MLX

Qwen 3.8 27B MLX-Quants (Vision, Text-Only & MTP)

Collection

Complete quantized MLX conversion of Qwen 3.8 27B. This collection contains vision-capable, text-only and all MTP quants. • 19 items • Updated Aug 15 • 4