no OS · straight to the metal

InferosBare-metal AI inference on the edge.

Inferos takes over the hardware and runs models directly on it — no operating system, no runtime. It boots, drives the GPU, and serves normal APIs, on a Raspberry Pi or an x86 mini-PC.

No OS. No runtime.
Nothing between your model and the metal.
// the machine boots, and the next thing it does is inference.
01

Boot to the metal

UEFI or bare-metal handoff. Inferos takes the CPU, starts the other cores, and owns the machine. There is nothing underneath it.

02

Bring up the hardware

Its own USB, FAT32, framebuffer and networking — drivers written from scratch for each board, and GPU kernels hand-written for the Intel Gen11 and VideoCore VI.

03

Run models on-device

Transformers and diffusion, in fp16 and int8, accelerated on the integrated GPU. Talk to it over an OpenAI-compatible API, or from the on-screen console.

Runs on the metal
Llama · Qwen3 · Gemma3 · SmolLM2 · LFM2 · Qwen3.5 · DDPM diffusion
10 flashable images · aarch64 & x86 · fp16 · int8 · OpenAI-compatible

Flash it. Boot it.

Download the Imager