500B+ Parameter LLMs. Locally. On a Mac Mini.

The most powerful AI shouldn't belong to a handful of corporations. Graviton is the open-source engine that breaks AI free from cloud monopolies and puts it on hardware you already own. One command. No cloud. No subscription. Just AI — for everyone.

Get Started View on GitHub

What Graviton Does

💻

72B Model on a Mac Mini

Models that corporations run on server farms — Graviton runs on your desk. A 72B model compressed from 144 GB to 36 GB, streamed layer by layer so it never exceeds your RAM.

🔓

Break Free From Cloud Monopolies

No OpenAI subscription. No API rate limits. No data harvesting. Your AI runs on your machine, owned by you, controlled by you.

🗜️

Shrink Models up to 10x

Graviton compresses AI models from 16-bit down to 4-bit or even 1.58-bit — making them 4 to 10 times smaller without losing the ability to think clearly.

🌊

Stream Models That Don't Fit in Memory

Even if a model is too large for your RAM, Graviton loads it one layer at a time from your SSD, compresses each layer instantly, and keeps going. The model never has to fit in memory all at once.

⚡

Fast Generation

Speculative decoding predicts multiple tokens at once and verifies them in a single step. Dynamic sparsity skips 70% of unnecessary computation per token. Result: 2–3x faster text generation.

🤖

Built for AI Agents Too

A headless API server with zero UI dependencies. AI agents on low-budget machines can load and use 70B+ models through a simple REST API — no GPU cluster, no cloud bill.

🧬

Graviton-Native: Efficient Architectures

Train models from scratch with BitNet (ternary weights) and MoE. 350M params in ~66 MB. 500B MoE with 32GB RAM. Technical report →

The Numbers

72B AI Model on a Mac

  • Without Graviton: 144 GB — crashes
  • With Graviton: 36 GB — runs on a laptop
  • Compression: 4x smaller
  • Hardware needed: Mac with 64 GB RAM
  • GPU server needed: No
  • Cloud subscription: No
  • Cost: Free

Get Started

For you — install and start chatting 
pip install graviton-ui && graviton-ui

Your browser opens. Pick a model. Start chatting. That's it.

For AI agents — headless API, zero UI 
pip install "graviton-ai[api]" && graviton-api

REST API on 0.0.0.0:7860. Load models, send messages, get streaming responses. No browser needed.

Read Graviton-Native Paper