228KB · 19 backends · BSL-1.1 · inference-x.com

Build your AI.
Own every layer of it.

AI models are sealed. You can't open them, modify them, or understand them.
Inference-X changes that. Decompose any model into its parts. Swap organs. Train adapters. Assemble exactly what you need. Deploy on your server. Keep your data.

See how it works
228KB
engine binary
19
hardware backends
68
deploy cities
models supported
// Architecture

A model is not a monolith.

Every AI model is made of distinct parts that can be separated, specialized, and recombined. We expose that structure so you can work with it directly.

# Before: the sealed monolith model.gguf # 70 billion params # one file, sealed # train everything # or change nothing # cost: millions # ────────────────── # After: open anatomy skeleton.bin # attention layers organ_domain.bin # specialized FFN adapter_lang.bin # LoRA personality # compose on demand
🦴
Skeleton
Attention layers — how the model thinks. Shared across builds.
REUSE
🧠
Organs
FFN layers — domain knowledge. Medical, legal, code, science.
TRAIN
Adapters
LoRA weights — personality, tone, language. Lightweight.
CUSTOMIZE
InferenceX Engine
228KB. Detects hardware. Runs on everything.
RUN
// Hardware-agnostic

The engine detects your hardware and adapts. Same binary everywhere.

💻
CPU x86
AVX2 · AVX512
🎮
NVIDIA GPU
CUDA 12+
🍎
Apple Silicon
Metal
🔴
AMD GPU
ROCm
ARM / Edge
NEON
🧮
Intel ARC
SYCL
📱
Mobile
Cortex
☁️
Cloud VPS
68 cities
// Plans

Build free. Scale when you need.

Every plan gives you access to the IX build platform. Paid plans unlock dedicated server instances deployed in the city nearest to you.

Tier 0
Micro
Free

Build and run open-source models. No credit card. Full engine access.

Shared nano · 1vCPU · 512MB
  • InferenceX engine (19 backends)
  • Open-source models
  • 3 AI configs
  • Local inference
  • Dedicated server
  • All models
  • Organ surgery
Tier 1
Studio
$49/month

Your own server. All models. Build, download, and share your AI.

Dedicated · 2vCPU · 2GB · 50GB NVMe
  • Everything in Micro
  • All models available
  • Unlimited builds
  • Download your model (GGUF)
  • 10K API calls/month
  • 68-city location choice
  • Organ surgery
~$5/mo infra · 90% margin
Tier 2
Atelier
$199/month

Full organ surgery. Train, graft, assemble. Publish to the IX Store and earn.

Dedicated · 4vCPU · 8GB · 125GB NVMe
  • Everything in Studio
  • Organ extract, graft, assemble
  • Train LoRA adapters
  • IX Store (keep 80%)
  • 100K API calls/month
  • White-label export
~$10/mo infra · 95% margin
Tier 3
Forge
Custom

Bare metal, GPU clusters, private deployment. For organizations that need full control.

Bare metal or multi-VPS · 32+ vCPU
  • Everything in Atelier
  • GPU clusters (A100/H100)
  • Private organ training
  • 99.9% SLA
  • Legal jurisdiction choice
  • Dedicated engineer
// IX Store

Share. Discover. Earn.

Publish specialized organs and adapters. Anyone can install them into their model. You keep 80% of every sale. Build once, earn forever.

Publish your own organs — Atelier plan required

// 68 cities

Deploy near your users.

Your instance is provisioned in the city closest to you — automatically. GDPR? Frankfurt. Data sovereignty? Fez. Latency? Singapore.

free
⬡ Your Instance
○ None
◈ My AIs
  • No builds yet
🔬 Organ View
skeleton.bin ← attention (shared) organ_domain.bin ← your knowledge adapter_lang.bin ← personality (LoRA)

Organ surgery on Atelier plan.

◉ Store

Publish organs, keep 80% of every sale.

$0
Revenue
0
Published