Reserve at $30 & Save Up to $2000!
Built by Developers. Tested in the Real World.
Built by Developers.
Tested in the Real World.
Specifications
Product Model
PineaPi Pro
PineaPi Lite
Positioning
Flagship desktop AI node — full local LLM + multimodal concurrent
AI-native workstation alternative — local AI deployment & desktop applications
Core Compute
AI Compute
275 TOPS (Int 8)
180 TOPS (Int 8)
GPU Compute
~170 TOPS GPU-class AI Acceleration, 2048 shader cores + tensor acceleration
Integrated graphics AI acceleration for lightweight generative inference
Compute Stack
CPU + GPU + dedicated AI accelerator unified compute pipeline
CPU + integrated graphics + NPU hybrid inference pipeline
AI Runtime
Optimized local AI inference runtime with graph-level acceleration
Cross-platform AI runtime with heterogeneous device scheduling
Max Model Scale
Up to 120B-class MoE LLM, quantized sparse-active inference
Up to 30B-class quantized LLM + multimodal workloads
Memory & Storage
Memory
64 GB LPDDR5
64 GB LPDDR5x
Storage
64 GB eMMC + M.2 NVMe slot for expansion
1 TB NVMe SSD (onboard)
Camera & Audio
Camera
1080P FHD board camera (face recognition / gesture / object detection / spatial awareness)
None
Microphone
6-mic circular array (far-field)
4-mic linear array (far-field)
Speaker
Stereo speaker
Built-in stereo speaker
Interface & Controls
Light Ring
Customizable Status LED
Customizable Status LED
AI Button
One-tap local/cloud model switching + voice input control
One-tap local/cloud model switching + voice input control
Wireless
WiFi 6 + Bluetooth 5.0
WiFi 6 + Bluetooth 5.2
Wired Network
1× Gigabit Ethernet
2× Gigabit Ethernet
USB Ports
USB-A ×2 + USB-C ×2
USB-A ×4 + USB-C ×2
Video Output
HDMI ×1
HDMI ×1 + DisplayPort ×1
OS & Software
OS
Ubuntu 22.04
Ubuntu 24.04
Pre-installed AI
MiniCPM series — text / multimodal / omni models ready out of box
MiniCPM series — text / multimodal / omni models ready out of box
Text-to-Speech
VoxCPM2 - 30 languages +
VoxCPM2 — 30 languages + 9 Chinese dialects
Agent Framework
ModelBest Proprietary AI Framework — pre-configured, ready out of the box
ModelBest Proprietary AI Framework — pre-configured, ready out of the box
Power & Cooling
Cooling
Active cooling
Active cooling
Power Consumption
15 W – 60 W
90 W
Power Input
28V Type-C (standard PD protocol)
DC 19V / 6A
Physical
Dimensions
120 × 120 × 75 mm
147 × 147 × 45 mm
Weight
~1090 g
~1000 g