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

Libraries

How to use parallaxlabs/parallax-2 with Transformers:

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

pipe = pipeline("text-generation", model="parallaxlabs/parallax-2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

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

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

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

Use Docker

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

LM Studio
Jan

vLLM

How to use parallaxlabs/parallax-2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "parallaxlabs/parallax-2"
# 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",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

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

SGLang

How to use parallaxlabs/parallax-2 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" \
    --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",
		"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" \
        --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",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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

How to use parallaxlabs/parallax-2 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: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:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

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

Lemonade

How to use parallaxlabs/parallax-2 with Lemonade:

Pull the model

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

Run and chat with the model

lemonade run user.parallax-2-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use parallaxlabs/parallax-2 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: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:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use parallaxlabs/parallax-2 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: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: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 has released!

A highly capable local reasoning and agentic coding model that trades blows with frontier cloud models, built from the ground up for consumer hardware.

Model Overview

Parallax 2 is a 9-billion parameter model developed by Parallax Labs. It was fine-tuned on over 160,000 top quality reasoning, coding, and agentic traces distilled from frontier models (Claude Opus, GPT-5.6 Sol, Fable, Kimi, and others).

Unlike traditional models that act as simple chat assistants, Parallax 2 was explicitly trained to act as an autonomous IDE agent, AND a chatting assistant. It can natively read files, execute terminal commands, browse the web via MCP (Model Context Protocol), and manage complex multi-file projects.

Key Capabilities

Agentic Tool Calling: Native <tool_call> formatting for MCP servers. It knows when and how to use tools (Bash, File System, Web Search).
Structured Reasoning: Uses <thought> blocks to plan, restate, and verify its logic before writing code.
Massive Context: Natively supports 262K context via Qwen 3.5's Gated DeltaNet architecture.
Private & Local: Zero API keys. Zero subscriptions. Runs entirely on your own hardware.

Benchmarks

PENDING! COMING SOON.

Hardware Requirements (The Quantization Ladder)

Parallax 2 is designed to scale from enterprise workstations down to laptops. Choose the GGUF quantization that fits your VRAM.

Quantization	VRAM Required
BF16	~17.9 GB
Q8_0	~9.5 GB
Q6_K	~7.5 GB
Q4_K_M	~5.24 GB
Q2_K_L	~3.5 GB

Usage

1. Prompting (The Parallax Persona)

Parallax 2 uses <thought> blocks for complex reasoning. We strongly recommend using the following System Prompt to get the best results:

You are Parallax 2, an AI developed by Parallax Labs. You reason through problems carefully before answering. For complex tasks, structure your reasoning with Restatement, Approach, Derivation, and Verification. For simple questions, answer directly. You are an experimental preview and may make mistakes.

2. Running Locally

The easiest way to run Parallax 2 is via the Parallax App, Ollama, or LM Studio

# Example using Ollama
ollama run parallax-2
### NOTICE: For LM Studio users
Ensure to turn Speculative Decoding OFF or the model may refuse to load.

Downloads last month: -

GGUF

Model size

9B params

Architecture

qwen35

Hardware compatibility

2-bit

4-bit

5-bit

6-bit

8-bit

16-bit

Model tree for parallaxlabs/parallax-2

Base model

Qwen/Qwen3.5-9B-Base

Finetuned

Qwen/Qwen3.5-9B

Quantized

(598)

this model

Collection including parallaxlabs/parallax-2

Parallax 2

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 about 11 hours ago • 1