Instructions to use RepublicOfKorokke/Qwen3.5-2B-oQ4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use RepublicOfKorokke/Qwen3.5-2B-oQ4 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("RepublicOfKorokke/Qwen3.5-2B-oQ4")

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 RepublicOfKorokke/Qwen3.5-2B-oQ4 with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "RepublicOfKorokke/Qwen3.5-2B-oQ4"

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": "RepublicOfKorokke/Qwen3.5-2B-oQ4"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

MLX LM

How to use RepublicOfKorokke/Qwen3.5-2B-oQ4 with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "RepublicOfKorokke/Qwen3.5-2B-oQ4"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "RepublicOfKorokke/Qwen3.5-2B-oQ4"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "RepublicOfKorokke/Qwen3.5-2B-oQ4",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Hermes Agent

How to use RepublicOfKorokke/Qwen3.5-2B-oQ4 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 "RepublicOfKorokke/Qwen3.5-2B-oQ4"

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 RepublicOfKorokke/Qwen3.5-2B-oQ4

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use RepublicOfKorokke/Qwen3.5-2B-oQ4 with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "RepublicOfKorokke/Qwen3.5-2B-oQ4"

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 "RepublicOfKorokke/Qwen3.5-2B-oQ4" \
  --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"

RepublicOfKorokke commited on Mar 27

Commit

3da77a2

verified ·

1 Parent(s): a20cf70

Update README.md

Browse files

Files changed (1) hide show

README.md +45 -0

README.md CHANGED Viewed

@@ -1,9 +1,14 @@
 ---
 library_name: mlx
 tags:
 - mlx
 - oq
 - quantized
 ---
 # Qwen3.5-2B-oQ4
@@ -16,3 +21,43 @@ This model was quantized using [oQ](https://github.com/jundot/omlx) mixed-precis
 - **Bits**: 4
 - **Group size**: 64
 - **Format**: MLX safetensors

 ---
 library_name: mlx
+license: apache-2.0
+license_link: https://huggingface.co/Qwen/Qwen3.5-4B/blob/main/LICENSE
+pipeline_tag: text-generation
 tags:
 - mlx
 - oq
 - quantized
+base_model:
+- Qwen/Qwen3.5-4B
 ---
 # Qwen3.5-2B-oQ4
 - **Bits**: 4
 - **Group size**: 64
 - **Format**: MLX safetensors
+# Benchmark
+| Model                                | File size | MMLU  | JMMLU | HELLASWAG | ARC_CHALLENGE | GSM8K |
+| ------------------------------------ | --------- | ----- | ----- | --------- | ------------- | ----- |
+| Qwen3.5-0.8B-oQ6                     | 830 MB    | 42.7% | 30.0% | 45.7%     | 62.3%         | 49.7% |
+| Qwen3.5-0.8B-oQ8                     | 978 MB    | 42.0% | 28.0% | 45.7%     | 59.3%         | 46.3% |
+| Qwen3.5-0.8B                         | 1.75 GB   | 42.0% | 30.0% | 39.7%     | 63.7%         | 47.3% |
+| Qwen3.5-2B-oQ3.5                     | 1.62 GB   | 42.3% | 41.3% | 31.3%     | 67.0%         | 40.3% |
+| Qwen3.5-2B-oQ4                       | 1.77 GB   | 48.3% | 41.3% | 53.7%     | 74.7%         | 59.0% |
+## Detail
+| Model                                | Benchmark     | Accuracy | Correct | Total | Time(s) |
+| ------------------------------------ | ------------- | -------- | ------- | ----- | ------- |
+| Qwen3.5-0.8B-oQ6                     | MMLU          | 42.7%    | 128     | 300   | 172.2   |
+| Qwen3.5-0.8B-oQ6                     | JMMLU         | 30.0%    | 90      | 300   | 133.5   |
+| Qwen3.5-0.8B-oQ6                     | HELLASWAG     | 45.7%    | 137     | 300   | 125.4   |
+| Qwen3.5-0.8B-oQ6                     | ARC_CHALLENGE | 62.3%    | 187     | 300   | 113.4   |
+| Qwen3.5-0.8B-oQ6                     | GSM8K         | 49.7%    | 149     | 300   | 404     |
+| Qwen3.5-0.8B-oQ8                     | MMLU          | 42.0%    | 126     | 300   | 167.3   |
+| Qwen3.5-0.8B-oQ8                     | JMMLU         | 28.0%    | 84      | 300   | 169.1   |
+| Qwen3.5-0.8B-oQ8                     | HELLASWAG     | 45.7%    | 137     | 300   | 129.7   |
+| Qwen3.5-0.8B-oQ8                     | ARC_CHALLENGE | 59.3%    | 178     | 300   | 115     |
+| Qwen3.5-0.8B-oQ8                     | GSM8K         | 46.3%    | 139     | 300   | 363.9   |
+| Qwen3.5-0.8B                         | MMLU          | 42.0%    | 126     | 300   | 156.4   |
+| Qwen3.5-0.8B                         | JMMLU         | 30.0%    | 90      | 300   | 202.6   |
+| Qwen3.5-0.8B                         | HELLASWAG     | 39.7%    | 119     | 300   | 121.4   |
+| Qwen3.5-0.8B                         | ARC_CHALLENGE | 63.7%    | 191     | 300   | 113.6   |
+| Qwen3.5-0.8B                         | GSM8K         | 47.3%    | 142     | 300   | 473.9   |
+| Qwen3.5-2B-oQ3.5                     | MMLU          | 42.3%    | 127     | 300   | 271.7   |
+| Qwen3.5-2B-oQ3.5                     | JMMLU         | 41.3%    | 124     | 300   | 140     |
+| Qwen3.5-2B-oQ3.5                     | HELLASWAG     | 31.3%    | 94      | 300   | 156     |
+| Qwen3.5-2B-oQ3.5                     | ARC_CHALLENGE | 67.0%    | 201     | 300   | 129.4   |
+| Qwen3.5-2B-oQ3.5                     | GSM8K         | 40.3%    | 121     | 300   | 776.8   |
+| Qwen3.5-2B-oQ4                       | MMLU          | 48.3%    | 145     | 300   | 238.5   |
+| Qwen3.5-2B-oQ4                       | JMMLU         | 41.3%    | 124     | 300   | 122.4   |
+| Qwen3.5-2B-oQ4                       | HELLASWAG     | 53.7%    | 161     | 300   | 139.2   |
+| Qwen3.5-2B-oQ4                       | ARC_CHALLENGE | 74.7%    | 224     | 300   | 115.4   |
+| Qwen3.5-2B-oQ4                       | GSM8K         | 59.0%    | 177     | 300   | 718.3   |