Instructions to use dylpickle13/Qwen3.5-2B-oQ8-mtp with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dylpickle13/Qwen3.5-2B-oQ8-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("dylpickle13/Qwen3.5-2B-oQ8-mtp")
config = load_config("dylpickle13/Qwen3.5-2B-oQ8-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 dylpickle13/Qwen3.5-2B-oQ8-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 "dylpickle13/Qwen3.5-2B-oQ8-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": "dylpickle13/Qwen3.5-2B-oQ8-mtp"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent

How to use dylpickle13/Qwen3.5-2B-oQ8-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 "dylpickle13/Qwen3.5-2B-oQ8-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 dylpickle13/Qwen3.5-2B-oQ8-mtp

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use dylpickle13/Qwen3.5-2B-oQ8-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 "dylpickle13/Qwen3.5-2B-oQ8-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 "dylpickle13/Qwen3.5-2B-oQ8-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"

Qwen3.5-2B-oQ8-mtp / PUBLISH_CHECK.md

dylpickle13

Mark clean verification download removed

941eccb verified 4 months ago

preview code

Raw

History Blame Contribute Delete

3.38 kB

Publish Check: Qwen3.5-2B-oQ8-mtp

Model

Published repo: dylpickle13/Qwen3.5-2B-oQ8-mtp
Base model: Qwen/Qwen3.5-2B
Base revision: 15852e8c16360a2fea060d615a32b45270f8a8fc
License: Apache-2.0
Runtime target: oMLX / MLX
Quantization/conversion: oMLX/oQ level 8, affine 8-bit, group size 64, bfloat16 preserved tensors, text_only=false, preserve_mtp=true

Local artifact verification

Local source path: /Users/dylanrapanan/.omlx/models/Qwen3.5-2B-oQ8-mtp
Required files present: PASS after adding upstream LICENSE
model.safetensors size: 2,735,417,896 bytes
safetensors metadata: {"format": "mlx"}
tensor count: 1020
MTP tensor count: 29
vision tensor count: 297
oMLX discovery before upload: model_type=vlm, mtp_compatible=true

Local oMLX smoke tests

oMLX server/version: local oMLX server; mlx-lm 0.31.3, mlx-vlm 0.5.0
loaded model name: Qwen3.5-2B-oQ8-mtp
text smoke test: PASS — prompt Reply with exactly OK. returned OK
image smoke test: PASS — generated red square returned red
notes: low-latency camera profile applied with 8192 context, 512 max tokens, thinking disabled, mtp_enabled=true, vlm_mtp_enabled=false

Private Hugging Face upload

private repo created: PASS
uploaded repo: dylpickle13/Qwen3.5-2B-oQ8-mtp
initial private upload commit: e5394295ebc0195d3944cc416d56c7a1d8de4b3d
license-added private commit: 65e35d8d606f85dd1c1f45f88b2035ab7ea7a474
uploaded files:
- .gitattributes
- README.md
- LICENSE
- chat_template.jinja
- config.json
- merges.txt
- model.safetensors
- preprocessor_config.json
- tokenizer.json
- tokenizer_config.json
- vocab.json

Clean-download verification

clean download path: projects/omlx/quant-publisher/downloads/Qwen3.5-2B-oQ8-mtp-hf-clean
file list verified: PASS
safetensors opens: PASS
tensor counts match: PASS
clean download tensor count: 1020
clean download MTP tensor count: 29
clean download vision tensor count: 297
oMLX clean load: PASS — local oMLX copy was deleted, then re-downloaded from the private Hugging Face repo, discovered, loaded, and tested
text smoke test: PASS — OK
image smoke test: PASS — red

Public release

repo made public: PASS
anonymous access verified: PASS — public API and raw README returned HTTP 200
final public URL: https://huggingface.co/dylpickle13/Qwen3.5-2B-oQ8-mtp
public release commit verified: 18207548b8ce5bc8d699d35cbe9dcf73d0bbd056
final verification note: this PUBLISH_CHECK.md was updated after public release, so the repository latest commit may be newer than the public release commit above.

Backup

Google Drive backup: NA for this run unless requested separately
backup folder URL: NA
checksum notes: safetensors metadata and tensor counts recorded above

Cleanup

old local Qwen3.5-0.8B-oQ8-mtp removed from oMLX: PASS
temporary base source Qwen/Qwen3.5-2B removed from oMLX: PASS
clean download removed: PASS
local project copy removed or retained intentionally: weights kept only under ignored downloads/ temporarily for verification; no weights committed