Your AI belongs to You

Spica is a local-first AI operating system that runs intelligent AI agents at breakthrough speeds on your own device.

Your data. Your OS. Your machine. Your AI.

Download Spica

Spica · Preview1:36

A film of Spica's interface, in development.

A 96-second film. It opens on a laptop: “A preview into the future of personal AI.” An amber ring over a lensed black hole, captioned that the AI runs on the device and nothing it sees or hears goes to the cloud. Over a dark desktop, the assistant, an amber ring, talks with its owner: offers to delete a 1:38 a.m. draft message, frees up the morning, points out a haircut in an hour and the vet at four, moves the haircut into an opening, and orders a bagel from Chicago Bagel that arrives in thirty minutes, while captions note that it sees the screen and application data in real time and that background agents handle busywork, all on the device. A lock screen reads 09:03, Tuesday, November 10, with the bagel at the door and a note that it keeps learning from the morning. Closing lines: “Your AI should be tailored to you.” “Your data should live on your device, not in someone else's datacenter.” It ends on “Spica. Your data. Your OS. Your AI.” by Micro Perceptron.

01 · localYour AI runs here, not in a data center.

On-device AI: models run on the machine in front of you — encrypted at rest, working with no connection at all. Your prompts, your files, and everything the system learns about how you work never need to leave it. Cloud models are a choice you make, never a dependency you inherit.

02 · secureSecure by construction.

Everything on your machine is backed by capabilities — cryptographic tokens that cannot be forged — protecting your computer from AI-backed cyberattacks.

03 · fastEvery chip you own, working at once.

A heterogeneous scheduler optimally uses every chip on your computer to run your AI at the speeds you need.

Supercharge your AI

Run local AI at breakthrough speeds

Spica is designed to optimize compute, memory, and disk speeds, achieving state-of-the-art AI inference speeds. Intel and AMD x86 devices are currently supported. Snapdragon (ARM) and RISC-V device support is coming soon. LLMs run on-device, 2.8× to 7.3× faster than llama.cpp on Linux on the same laptops.

We tested the latest Qwen models on consumer-grade laptops — see the benchmarks for every comparison.

Left, the Spica chat running QUASAR-QAT/Qwen3.8-27B-QUASAR-NVFP4 at Xhigh effort: a Python test harness for a matmul function, then its actual output, the 2×2 product [[19, 22], [43, 50]], a 3×3 identity check and an associativity check both True, and timings for sizes 1 to 100 rising from 0.0 to 133.3 ms, followed by observations that the checks passed and the scaling is cubic. The status line reads 1039 tokens at 23.4 tok/s, context 5281 of 187392, 5.41 tokens a step. Right, a terminal running kore status mem: 26.7 of 30.8 GiB of physical memory in use, page classes of 4 KiB, 2 MiB and 1 GiB, and DRAM stream bandwidth per core class.(open full size)
Local model + Python toolMSI Prestige 16 AI+ C3MTG · Core Ultra X7 358H · 32 GB

Qwen3.8-27B writes matrix multiplication in Python, runs it with its own Python tool and checks it: a known product, the identity, associativity and cubic scaling. 23.4 tok/s with a 187K-token context, and no cloud anywhere in the loop.

Vision

Every operating system running today was built before AI could run on the machine itself.

So AI runs in someone else's data center, and your data goes along with every request. Your OS at home isn't designed to run fast AI.

01Old foundations

Decades of runway, none of it aimed at this.

Every popular OS predates AI by a generation. Everything was designed for a person clicking through files and apps, not for AI agents that read, decide, and act on their own.

1991
Linux kernel released
1993
Windows NT architecture
2001
Mac OS X, same Unix lineage
22 yearsnothing AI-shaped yet
2023–25
AI becomes an everyday workload — mostly rented from the cloud
Now
Still running on the same foundations
02Underused hardware

The hardware is already there.

Most machines sold in the last couple of years already have a CPU, a GPU, and an NPU — a dedicated AI chip built specifically for matrix math. They could run capable models locally, but most software still only ever asks one of them to do anything. When that's not enough, they ship the work to the cloud instead while your machine sits idle.

Most software, today
CPU
GPU
NPU
Spica
CPU
GPU
NPU

Spica treats all three as one machine to schedule across, not three separate afterthoughts — which is what makes a capable model practical on a laptop in the first place.

03One system

Three parts, one experience.

Local AI that works the moment it's plugged in.

A frame from our film of Spica's interface, in development: its amber ring glows on a dark screen above the words “Morning! How are we doing?”

In development

SpicaThe operating system

Your personal AI assistant. Spica handles your day to day tasks, and constantly learns about you.

Picotron, Kore's mascot: a retro CRT monitor wearing pixel-art sunglasses

Running today

KoreThe kernel

Our custom capability microkernel. Built for AI.

How Kore works
A render of the Stone: a round, matte black computer with a glass top, an amber light ring around its edge and the Kore mascot engraved in outline on the glass.

In development

StoneThe computer

An AI PC built to run Spica, with an Intel Core Ultra X7 and up to 64 GB of memory.

Meet the Stone
04The goal
We are the gold standard operating system for local AI.

Fast, secure, and private AI should belong to everybody. You are paying for your hardware, so you should get to use it.

Download Spica