Model Zoo.

A comprehensive registry of intelligence kernels and cloud APIs officially supported by Luca OS. Local models execute entirely on-device with zero data telemetry. Cloud models require network access and API keys.

Brain / Core Logic

🧠
Local

Qwen 2.5

Alibaba • 7B Parameters
Context32k
Size4.2GB
VRAM6GB
🐋
Local

DeepSeek R1

DeepSeek • 8B Parameters
Context128k
Size4.7GB
VRAM8GB
🦙
Local

Llama 3.2

Meta AI • 3B Parameters
Context8k
Size2.0GB
VRAM4GB
🧠
Cloud API

Claude 3.5 Sonnet

Anthropic • Frontier
Context200k
LatencyLow
ModalityCoding Focus
⚡
Cloud API

GPT-4o

OpenAI • Frontier
Context128k
LatencyUltra-Low
ModalityMultimodal

Vision & Parsing

👁️
Local

UI-TARS Vision

Tencent • 7B Parameters
RoleGUI Control
CoordinateNative
VRAM8GB
🖼️
Local

SmolVLM

HuggingFace • 2B Parameters
RoleFast Parsing
Context8k
VRAM3GB

Memory / Vectors

🧪
Rank 2

Model2Vec Potion

Local • 5MB
Size5 MB
Speed500x Faster
RoleMobile Vector
🍞
Rank 6

MixedBread XS

Local • 90MB
Size90 MB
QualityBalanced
RoleCore Memory
🗂️
Rank 7

BGE Small En

Local • 130MB
Size130 MB
MTEB62.17
RoleHigh Accuracy

Voice Engine

🎙️
TTS Rank 8

Kokoro-82M

TTS • 10 Voices
Size100 MB
RAM Req512 MB
RolePremium Voice
⚡
TTS Rank 1

Piper Amy

TTS • Ultra-Fast
Size60 MB
RAM Req256 MB
RoleZero-Latency
🚀
STT Rank 2

Moonshine Tiny

STT • Next-Gen
Size190 MB
Speed15x Whisper
RoleFast Listen

Ollama Native Bridge

Luca OS seamlessly interfaces with your existing Ollama installation. If you already have a model running locally (e.g. llama3.2, qwen2.5, or custom fine-tunes), Luca will dynamically detect it and bypass the standard download sequence, booting instantly using your existing hardware-accelerated endpoints.

Auto-Discovery Port 11434
// Startup Initialization
[LUCA] Initiating hardware scan...
[LUCA] Polling existing inference engines
[LUCA] Found endpoint localhost:11434
[LUCA] Discovered 'qwen2.5:latest'
[LUCA] Bypassing native provision. Booting OS...