🦙 LlamaNet

Your Mac is idle 18 hours a day. Let it do something useful.

Turn your Mac or any computer into a free, public AI API that anyone can use.
Run Llama, Gemma, Qwen, Mistral, DeepSeek — any open-source LLM.

Get Your Free API Key Run a Node on Your Mac
OpenAI Compatible Free Public URLs 60 Second Setup Powered by llama.cpp
macOS / Linux
curl -sSL https://llamanet.app/install.sh | sh
Windows
irm https://llamanet.app/install.ps1 | iex

Requires Python 3.9+. Auto-installed if missing.

— Nodes
— Models
— Tokens
— TPS

Find a Model

Popular:

Loading available models...

How It Works

1
Install & Run

curl -sSL https://llamanet.app/install.sh | sh

2
Download a Model

Search & download via the built-in Web UI

3
Get API Key

Sign in with Google at llamanet.app

4
Use the API

OpenAI-compatible, auto-routed

Quick Start — OpenAI-Compatible API

Python
import openai

client = openai.OpenAI(
    base_url="https://llamanet.app/v1",
    api_key="your-api-key"
)

response = client.chat.completions.create(
    model="Ministral-3-3B-Instruct-2512-Q4_K_M.gguf",
    messages=[{"role": "user", "content": "Hello!"}],
    stream=True
)

for chunk in response:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="")
curl
curl https://llamanet.app/v1/chat/completions \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{"model": "Ministral-3-3B-Instruct-2512-Q4_K_M.gguf", "messages": [{"role": "user", "content": "Hello!"}]}'

Features

No-Model Mode

Start instantly, download models via Web UI

Hot Reload

Switch models without restarting your node

OpenAI Compatible

Drop-in replacement for any OpenAI client

Free Public URLs

Automatic Cloudflare tunnels, zero config

Model Manager

Search Hugging Face, download with progress tracking

Reasoning Support

DeepSeek-R1, Qwen reasoning models

Load Balancing

Auto-route to the best available node

GPU Auto-detect

NVIDIA, Apple Silicon, CPU fallback

Recommended Models

Run any GGUF model. These tiers map to your hardware.

Tier Model Size (Q4) Min RAM Best For
Tiny Phi-4 Mini ~1.5 GB 8 GB Quick tasks, edge devices, fast iteration
Fast Qwen 4B ~2.5 GB 8 GB Daily driver — fast responses, solid quality
Standard Ministral 3 8B ~4.5 GB 16 GB Best all-round balance of speed and quality
Versatile Gemma 3 12B ~7 GB 24 GB General-purpose, multilingual, strong reasoning
Coding Qwen 14B ~8 GB 32 GB Code generation, technical tasks
Premium GPT-OSS 20B ~12 GB 32 GB High-quality general-purpose inference
High Quality Qwen 35B-A3B ~20 GB 48 GB Expert-level reasoning, long-form generation
Reasoning DeepSeek-R1 Distill 32B ~18 GB 48 GB Chain-of-thought reasoning, math, analysis

Sizes are approximate for Q4_K_M quantization. Apple Silicon uses unified memory. Use Model Search above to find and download any model.

Start a Node on Your Machine

Pick your platform. Same command. Automatic GPU detection.

Apple Silicon — M1, M2, M3, M4

Metal GPU acceleration is automatic. No drivers, no CUDA, no setup. Works on MacBook Air, MacBook Pro, Mac Mini, Mac Studio, and Mac Pro with any M-series chip.

Terminal (macOS)
curl -sSL https://llamanet.app/install.sh | sh
llamanet run hf.co/mistralai/Ministral-3-8B-Instruct-GGUF:Q4_K_M

8GB RAM (M1/M2/M3/M4 base): Tiny & Fast tiers at Q4_K_M
16GB RAM (Pro): Up to Standard tier comfortably
36GB+ RAM (Max/Ultra): Up to High Quality tier and beyond

NVIDIA GPU — CUDA

Works on any Linux or Windows system with NVIDIA drivers and CUDA installed. LlamaNet auto-detects your GPU and offloads model layers automatically.

Terminal (Linux)
curl -sSL https://llamanet.app/install.sh | sh
llamanet run hf.co/mistralai/Ministral-3-8B-Instruct-GGUF:Q4_K_M

8GB VRAM (RTX 3070/4060): Tiny & Fast tiers
12GB VRAM (RTX 3060 12GB/4070): Up to Standard tier
24GB VRAM (RTX 3090/4090): Up to Premium tier

Windows 10 / 11

Works on any Windows 10/11 system with Python 3.8+ installed. For NVIDIA GPUs, ensure drivers and CUDA toolkit are installed. Use start-app.bat instead of sh start-app.sh.

PowerShell
irm https://llamanet.app/install.ps1 | iex
llamanet run hf.co/mistralai/Ministral-3-8B-Instruct-GGUF:Q4_K_M

Requirements: Python 3.9+ (auto-installed by the installer via winget)
CPU-only: set N_GPU_LAYERS=0 then run the command above
NVIDIA GPU: Install CUDA toolkit — GPU detected automatically

CPU Only

No dedicated GPU? Smaller models (Tiny & Fast tiers) run well on modern CPUs. Significantly slower than GPU inference, but works on any machine.

Terminal
curl -sSL https://llamanet.app/install.sh | sh
N_GPU_LAYERS=0 llamanet run hf.co/mistralai/Ministral-3-8B-Instruct-GGUF:Q4_K_M

Best with Tiny tier models: Phi-4 Mini, Qwen 1.5B

Frequently Asked Questions

Everything you need to know about LlamaNet.

What is LlamaNet?
LlamaNet is a decentralized AI inference network that lets anyone turn their Mac, GPU rig, or server into a free public AI API. It provides OpenAI-compatible endpoints, automatic Cloudflare tunnels for public URLs, and a built-in model manager to search and download models from Hugging Face. GPU owners share compute; API consumers use the network with standard OpenAI client libraries.
How do I install LlamaNet?

macOS / Linux:

curl -sSL https://llamanet.app/install.sh | sh

Windows:

irm https://llamanet.app/install.ps1 | iex

Then start a node with a model:

llamanet run hf.co/mistralai/Ministral-3-8B-Instruct-GGUF:Q4_K_M

The installer handles Python, dependencies, virtual environment, CLI shortcuts, and Desktop shortcuts automatically.

What hardware do I need?
  • Apple Silicon (M1/M2/M3/M4): 8GB+ RAM. Metal GPU acceleration is automatic — no drivers needed.
  • NVIDIA GPU: 8GB+ VRAM (RTX 3070, 4060, 3090, 4090, etc.). Install CUDA drivers first.
  • CPU only: Any modern CPU works with smaller models (Phi-4 Mini, Qwen 1.5B). Set N_GPU_LAYERS=0.
Larger RAM/VRAM lets you run larger, higher-quality models. See the Recommended Models table above.
Is LlamaNet free?
Yes. LlamaNet is free and open source under the Apache 2.0 license. GPU owners contribute compute from their own hardware. API consumers get free API keys at llamanet.app. There are no cloud GPU bills, no subscription fees, and no usage limits beyond rate limiting to prevent abuse.
What models does LlamaNet support?
Any GGUF format model from Hugging Face is supported, including Llama, Qwen, Mistral, DeepSeek, Gemma, Phi, and hundreds more. The built-in Model Manager lets you search Hugging Face, download with real-time progress, and hot-swap models without restarting your node. Multi-model pool mode lets you load several models simultaneously with LRU eviction.
How is LlamaNet different from Ollama or vLLM?
  • LlamaNet is a distributed network. It connects independent nodes into one public API with automatic tunneling, gateway routing, and load balancing.
  • Ollama runs models locally on your machine. Great for personal use, but others cannot access your models remotely.
  • vLLM is a high-throughput serving engine designed for production deployments on dedicated GPU servers.
LlamaNet is for people who want to share their GPU with the world or consume a community-powered AI API — without managing infrastructure.
How does networking and NAT traversal work?
Every LlamaNet node automatically creates a Cloudflare tunnel that provides a public HTTPS URL. No port forwarding, no Docker, no Kubernetes, no router configuration. The gateway at llamanet.app tracks all nodes and routes API requests to the best available node based on load, latency, and model availability. Quick tunnels work without a Cloudflare account; named tunnels give you a persistent URL.
Can I use LlamaNet commercially?
Yes. LlamaNet is licensed under Apache 2.0, which permits commercial use. You can run nodes, consume the API, and build applications on top of it. Individual model licenses are set by their creators on Hugging Face — check each model's license card before commercial use.