Engine v3.0 — 19 backends, 305KB

Your Intelligence.
Your Machine.
Forever.

Build, run, and own your AI — on any hardware, in any language, with zero cloud dependency.

⚡ Build Your AI Explore Models →
305KBKB
Engine Size
7
Models
19
Backends
0€
Cost
$ ./inference-x qwen2.5-3b.gguf --serve 8080
═══ InferenceX v3.0 ═══
Model: Qwen 2.5 3B (Q4_K_M)
RAM: 2.1 GB / Backend: AVX2
Speed: 42 tok/s
✓ Listening on 0.0.0.0:8080
$ curl localhost:8080/v1/chat/completions
{"model":"qwen2.5-3b","choices":[...]}
OpenAI-compatible. Drop-in replacement.

Build Your AI

Tell us about your hardware. We recommend the perfect model.

Recommended

Confidence:

Intelligence Store

Community models, organs, and configurations. Free forever.

Loading...

Organ Transplant

Graft neural layers between models. Create hybrids that no one else has.

Compatible Layers

The Forge

Fine-tune any model on your data. Make it yours.

Loading...

Open Source

Every line of code. No secrets. Self-hosted.

Loading...
// Origin

origin_title

origin_sub

Why

Intelligence belongs to everyone. Not to corporations. Not to clouds. Not to subscriptions.

InferenceX is a universal compiler for AI models. 305KB. Runs everywhere — from a Raspberry Pi to a datacenter. 19 backends. One binary.

origin_quote

We were banned from GitHub. So we built our own. We don't depend on anyone's platform.

Protected by INPI patents. Open source under BSL 1.1.
origin_sig

🏔️ Morocco

origin_atlas

⚡ Engine Size

305 KB

origin_engine

🌍 Why

Mission: Make intelligence free, local, and permanent.

🧬 Organ Transplant

origin_organ