Almandine OS is a freestanding, zero Linux x86_64 operating system and local LLM inference appliance that boots directly from UEFI firmware into a live prompt in under 500 ms. Engineered for Intel hybrid silicon, it unlocks AVX2, FMA3, AVX VNNI, and High Performance hardware P-States in pure assembly, streaming multi gigabyte models from raw storage with zero kernel, daemon, or network overhead.
Almandine OS eliminates general purpose kernels, background daemons, swap partitions, network stacks, and bloated software layers, turning an x86_64 machine into a hardened offline model AI workstation.
Started in October 2026, Almandine OS is a compact operating system created to run local AI cleanly without the complexity of a traditional desktop distribution.
By removing general purpose kernels, background services, swap files, and network connections, Almandine OS acts as a private local AI workspace and a reliable offline AI appliance that boots from portable media whenever cloud services or external networks are unavailable.
A core mission of Almandine OS is building a highly secure operating system with minimal external dependencies. Instead of inheriting millions of lines of third party OS packages, shared libraries, and background daemons, the entire system compiles into a single auditable static executable binary.
With zero network drivers, zero open ports, zero dynamic library loading, and full physical RAM residency on dedicated Performance Cores, both runtime attack surface and software supply chain risk stay minimal by design.
Almandine OS boots straight from UEFI firmware into a single static executable binary, keeping the entire model resident in physical RAM with zero disk swap or background scheduler jitter.
On hybrid Intel Core processors, Almandine OS enables AVX2, FMA3, AVX VNNI, and High Performance hardware P-States across all dedicated Performance Cores while parking Efficiency Cores so 100% of power and memory bandwidth serves inference.
almandine> shell & Unnamed model live REPL
Almandine OS boots straight into an interactive command and AI prompt. Click any command button below or type your own question at the almandine> prompt to preview system commands and local Unnamed model replies.
Almandine OS keeps the entire Unnamed model in system memory for fast, responsive generation without disk slowdowns. Choose the model edition that matches your computer's available RAM.
| Requirement | Compact Edition (Unnamed model small) |
Standard Edition (Unnamed model medium) |
Full Appliance (Unnamed model large) |
|---|---|---|---|
| Model Size | ~2 GB | ~3 GB | ~16 GB |
| Minimum System RAM | 4 GB | 8 GB | 24 GB (32 GB recommended) |
| Processor Cores | 2 to 4 Physical Cores | 6 to 8 Physical Cores | 8 Performance Cores |
| Recommended CPU | Modern 64 bit Intel or AMD PC | Intel 12th Gen+ or AMD Ryzen | Intel Core i9 & Xeon |
| System Memory | DDR4 | DDR4 or DDR5 | Fast Dual Channel DDR5 |
| Boot Drive Size | 4 GB+ USB Drive | 8 GB+ USB 3.0 Drive | 32 GB+ Fast USB 3.2 or SSD |
Select an Unnamed model edition or adjust the sliders to estimate minimum system memory needed.
.iso images
Prebuilt bootable .iso images bundle Almandine OS with local Unnamed model weights so you can write a USB drive and boot directly into an offline AI prompt. Official release images are coming soon.
almandine-os-v0.1-compact.iso
Lightweight bootable image designed for everyday laptops and compact desktops with 4 GB or more of system memory.
almandine-os-v0.1-standard.iso
Balanced local reasoning and fast token generation for modern 8 GB+ PCs and portable offline workstations.
almandine-os-v0.1-appliance.iso
Flagship full capability local AI appliance image tuned for Intel Core i9 and Xeon systems with 24 GB to 32 GB+ of fast memory.