Instructions to use parallaxlabs/parallax-2-mini-preview with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use parallaxlabs/parallax-2-mini-preview with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="parallaxlabs/parallax-2-mini-preview")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("parallaxlabs/parallax-2-mini-preview", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use parallaxlabs/parallax-2-mini-preview 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 parallaxlabs/parallax-2-mini-preview:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf parallaxlabs/parallax-2-mini-preview:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf parallaxlabs/parallax-2-mini-preview:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf parallaxlabs/parallax-2-mini-preview:Q4_K_M

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 parallaxlabs/parallax-2-mini-preview:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf parallaxlabs/parallax-2-mini-preview:Q4_K_M

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 parallaxlabs/parallax-2-mini-preview:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf parallaxlabs/parallax-2-mini-preview:Q4_K_M

Use Docker

docker model run hf.co/parallaxlabs/parallax-2-mini-preview:Q4_K_M

LM Studio
Jan

vLLM

How to use parallaxlabs/parallax-2-mini-preview with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "parallaxlabs/parallax-2-mini-preview"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "parallaxlabs/parallax-2-mini-preview",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/parallaxlabs/parallax-2-mini-preview:Q4_K_M

SGLang

How to use parallaxlabs/parallax-2-mini-preview with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "parallaxlabs/parallax-2-mini-preview" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "parallaxlabs/parallax-2-mini-preview",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "parallaxlabs/parallax-2-mini-preview" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "parallaxlabs/parallax-2-mini-preview",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use parallaxlabs/parallax-2-mini-preview with Ollama:
```
ollama run hf.co/parallaxlabs/parallax-2-mini-preview:Q4_K_M
```
Unsloth Desktop

How to use parallaxlabs/parallax-2-mini-preview with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf parallaxlabs/parallax-2-mini-preview:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "parallaxlabs/parallax-2-mini-preview:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use parallaxlabs/parallax-2-mini-preview with Docker Model Runner:
```
docker model run hf.co/parallaxlabs/parallax-2-mini-preview:Q4_K_M
```

Lemonade

How to use parallaxlabs/parallax-2-mini-preview with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull parallaxlabs/parallax-2-mini-preview:Q4_K_M

Run and chat with the model

lemonade run user.parallax-2-mini-preview-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use parallaxlabs/parallax-2-mini-preview with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf parallaxlabs/parallax-2-mini-preview:Q4_K_M

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 parallaxlabs/parallax-2-mini-preview:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use parallaxlabs/parallax-2-mini-preview with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf parallaxlabs/parallax-2-mini-preview:Q4_K_M

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 "parallaxlabs/parallax-2-mini-preview:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Parallax-2-Mini-Preview

Parallax-2-Mini-Preview is the smallest tier of the Parallax-2 family, developed by Parallax Labs.

It is a 0.8B parameter model designed to demonstrate the Parallax reasoning style on the absolute minimum hardware. It is small enough to run entirely inside a web browser tab via WebAssembly (wllama/WebLLM) on a standard Chromebook, requiring zero server backend, zero API keys, and zero data leaving the device.

This model is not intended for heavy workloads at all, it is designed for testing and demos, it is just 800 million parameters, which is too low for about any task beyond a "conversation".

Model Details

Model Description

Developed by: Parallax Labs (a subsidiary of Parallax Studios)
Model type: Causal Language Model (LoRA finetune)
Language(s) (NLP): English
License: Apache 2.0
Finetuned from model: Qwen 3.5-0.8B

Model Sources

Repository: parallax-labs/parallax-2-mini-preview
Demo: labs.parallaxstudios.xyz/chat/wch

Uses

Direct Use

Parallax-2-Mini-Preview is intended for:

Browser-based AI: Running locally via WebLLM or wllama on hardware with as little as 2GB of free RAM.
Structured reasoning demos: Demonstrating step-by-step math and logic formatting (## Step 1, ## Answer) on low-end devices.
Fast local inference: Hitting 100+ tokens per second on modest hardware.

Out-of-Scope Use

Because this is a 0.8B model, it has severe limitations. It is not intended for:

Deep logic or complex coding: It will confidently output incorrect Python scripts or fail multi-step logic puzzles.
Long context: Trained at 8K context; it degrades heavily past 4K.
General knowledge: It has very little world knowledge and will hallucinate facts outside its training distribution.

Bias, Risks, and Limitations

Overthinking (Reasoning Collapse): Because 55% of its training data was deep reasoning traces, the model has a tendency to apply complex reasoning structures to simple questions (e.g., writing a 5-step derivation for "Hello").
Factual Hallucinations: As an 0.8B model, it will make up facts. Do not use it for factual retrieval.
Formatting quirks: It occasionally generates Markdown headers (###) and bold text inappropriately for casual conversation.

Recommendations

Treat this model as a format demonstrator, not an oracle. Use the provided system prompt to keep it on track:

"You are Parallax-2-Mini-Preview. You are a helpful, smart assistant. If asked a complex question, briefly show your reasoning step-by-step before giving a thorough, detailed answer. Do not over-explain simple facts."

How to Get Started with the Model

In the Browser (Zero Install)

You can run this model entirely in your browser using the wllama integration on our website: labs.parallaxstudios.xyz/chat/wch

Via llama.cpp / LM Studio (GGUF)

Download the Q4_K_M GGUF file and load it into your favorite local inference app.

Downloads last month: 59

GGUF

Model size

0.8B params

Architecture

qwen35

Hardware compatibility

4-bit

Model tree for parallaxlabs/parallax-2-mini-preview

Base model

Qwen/Qwen3.5-0.8B-Base

Finetuned

Qwen/Qwen3.5-0.8B

Quantized

(358)

this model

Datasets used to train parallaxlabs/parallax-2-mini-preview

Collection including parallaxlabs/parallax-2-mini-preview

Parallax 2 Family

Collection

Introducing the Parallax 2 family of AI models, a modern family of AI models based on Qwen, designed for powerful local reasoning and coding. • 4 items • Updated 1 day ago • 1