Instructions to use catblob-hash/Mewrk-Qwen3.5-0.8B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use catblob-hash/Mewrk-Qwen3.5-0.8B with MLX:

# Download the model from the Hub
pip install huggingface_hub[hf_xet]

hf download catblob-hash/Mewrk-Qwen3.5-0.8B --local-dir Mewrk-Qwen3.5-0.8B

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use catblob-hash/Mewrk-Qwen3.5-0.8B with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf catblob-hash/Mewrk-Qwen3.5-0.8B:F16
# Run inference directly in the terminal:
llama cli -hf catblob-hash/Mewrk-Qwen3.5-0.8B:F16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf catblob-hash/Mewrk-Qwen3.5-0.8B:F16
# Run inference directly in the terminal:
llama cli -hf catblob-hash/Mewrk-Qwen3.5-0.8B:F16

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf catblob-hash/Mewrk-Qwen3.5-0.8B:F16
# Run inference directly in the terminal:
./llama-cli -hf catblob-hash/Mewrk-Qwen3.5-0.8B:F16

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf catblob-hash/Mewrk-Qwen3.5-0.8B:F16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf catblob-hash/Mewrk-Qwen3.5-0.8B:F16

Use Docker

docker model run hf.co/catblob-hash/Mewrk-Qwen3.5-0.8B:F16

LM Studio
Jan
Ollama
How to use catblob-hash/Mewrk-Qwen3.5-0.8B with Ollama:
```
ollama run hf.co/catblob-hash/Mewrk-Qwen3.5-0.8B:F16
```
Unsloth Desktop

How to use catblob-hash/Mewrk-Qwen3.5-0.8B with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "catblob-hash/Mewrk-Qwen3.5-0.8B"

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": "catblob-hash/Mewrk-Qwen3.5-0.8B"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use catblob-hash/Mewrk-Qwen3.5-0.8B with Docker Model Runner:
```
docker model run hf.co/catblob-hash/Mewrk-Qwen3.5-0.8B:F16
```

Lemonade

How to use catblob-hash/Mewrk-Qwen3.5-0.8B with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull catblob-hash/Mewrk-Qwen3.5-0.8B:F16

Run and chat with the model

lemonade run user.Mewrk-Qwen3.5-0.8B-F16

List all available models

lemonade list

Hermes Agent

How to use catblob-hash/Mewrk-Qwen3.5-0.8B 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 "catblob-hash/Mewrk-Qwen3.5-0.8B"

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 catblob-hash/Mewrk-Qwen3.5-0.8B

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use catblob-hash/Mewrk-Qwen3.5-0.8B with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "catblob-hash/Mewrk-Qwen3.5-0.8B"

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 "catblob-hash/Mewrk-Qwen3.5-0.8B" \
  --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"

Mewrk-Qwen3.5-0.8B

Builds of Qwen/Qwen3.5-0.8B, the official instruct model, at commit 2fc06364715b967f1860aea9cf38778875588b17, for the local helper model of Mewrk (conversation titles, read from the whole message with its attached files and images, and shell command explanations). Mewrk downloads one of them when the user turns the feature on; each file is pinned by size and SHA-256 in the app (catalog.json here is the same list).

Directory	Backend	Platform
`ane/qwen35-ane-2/`	Core ML package for the Apple Neural Engine (5,120 positions per sequence), compiled on the device after download	Macs with a Neural Engine, macOS 15+
`ane/qwen35-ane-1/`	The same at 1,024 positions, an older layout
`mlx/qwen35-mlx-1/`	MLX weights (page-aligned for zero-copy loading) for the GPU	Apple silicon, macOS 14+
`mlx/runtime/0.32.2/`	`mlx.metallib`, the kernels of MLX 0.32.2 (Apple's `mlx-metal` wheel, MIT, see its `LICENSE`)	with the MLX build
`vision/qwen35-vision-1/`	The vision tower, the checkpoint's `model.visual.*` tensors as they are (BF16 safetensors), run on the CPU	with the Core ML and MLX builds
`llama/qwen35-gguf-f16-1/`	GGUF (f16), byte for byte what llama.cpp's converter writes	llama.cpp: Windows, Linux
`llama/qwen35-mmproj-f16-1/`	Vision projector (GGUF, f16) for llama.cpp's `mtmd`, byte for byte what llama.cpp's converter writes with `--mmproj`	with the GGUF

These files are not general-purpose conversions: the Core ML package and the MLX weights have the layout Mewrk's own inference code expects.

How they are made

src-tauri/local-model/examples/build_release.rs in the Mewrk repository writes all of them from the official release. The output is deterministic, so anyone can rebuild and compare:

cargo run --release -p mewrk-local-model --features mlx --example build_release -- <Qwen3.5-0.8B dir> <out dir>

License

The model is Apache-2.0, as released by the Qwen team (LICENSE). The files here are converted from it (to float16 and to each backend's layout). mlx/runtime/0.32.2/mlx.metallib is part of MLX, MIT-licensed by Apple.

Downloads last month: 158

MLX

Hardware compatibility

Quantized

GGUF

Model size

0.8B params

Architecture

qwen35

Hardware compatibility

16-bit

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for catblob-hash/Mewrk-Qwen3.5-0.8B

Base model

Qwen/Qwen3.5-0.8B-Base

Finetuned

Qwen/Qwen3.5-0.8B

Quantized

(357)

this model