Specifications

Model

OrangePi O1

SoC

Sophgo BM1688

CPU

8x Arm Cortex-A53 up to 1.65 GHz

Co-processor

RISC-V C906 up to 1.2 GHz

AI Accelerator

Integrated TPU - 16 TOPS INT8 / 32 TOPS INT4

FP16 / FP32

4.0 TFLOPS at FP16/BF16, 0.5 TFLOPS at FP32

Memory

8GB or 16GB LPDDR4 / LPDDR4X

eMMC

Optional 16 / 32 / 64 / 128 GB modules

SPI Flash

Optional 128Mb or 256Mb (mutually exclusive with eMMC)

Mass Storage

M.2 M-Key PCIe 3.0 NVMe SSD + SATA 3.0 + microSD up to 128GB

Video Decode

H.264 / H.265 up to 16 simultaneous 1080p30 streams

Video Encode

H.264 / H.265 up to 10x 1080p30, plus JPEG encode/decode

Board Size

90 x 70 mm

AI Frameworks

Caffe, PyTorch, TensorFlow, ONNX, MXNet

Target Workloads

Large language models, vision-language models, image analysis, multi-object tracking

Vendor

Orange Pi

Overview

The OrangePi O1 is a 90 x 70 mm single board computer built around the Sophgo BM1688, an edge-AI SoC that pairs eight Arm Cortex-A53 cores with a RISC-V C906 co-processor and an integrated TPU. Sophgo rates the TPU at 16 TOPS INT8 peak, doubling to 32 TOPS when INT4 quantisation is used, with 4.0 TFLOPS at FP16/BF16.

What separates the O1 from a generic Arm SBC is storage and memory headroom for model work: up to 16GB of LPDDR4/LPDDR4X, an M.2 M-Key slot for a PCIe 3.0 NVMe SSD, a dedicated SATA 3.0 interface and up to 128GB of optional eMMC. Note that the eMMC and SPI-flash options are mutually exclusive on this board.

Video throughput is specified for multi-channel vision: H.264/H.265 decode of up to sixteen simultaneous 1080p30 streams and encode of up to ten, both with JPEG assistance. Sophgo lists Caffe, PyTorch, TensorFlow, ONNX and MXNet in the toolchain, and the target workloads are explicitly LLM and VLM inference, image analysis and multi-object tracking at the edge.

Key Benefits

16 TOPS INT8 TPU in a 90 x 70 mm board, up to 16GB LPDDR4X for larger models, M.2 NVMe plus SATA 3.0 for local model and dataset storage, and 16-channel 1080p30 decode for multi-camera vision.

Applications

Edge AI inference nodes, multi-camera video analytics, local LLM and VLM serving, industrial machine vision, and robotics perception platforms.

Request a Quote — ORANGEPI O1 (SOPHGO BM1688) — 16-TOPS EDGE AI SBC WITH SATA 3.0

QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.

Get Your Quote →

Related Products

Banana Pi BPI-CM7 (Allwinner H618) Compute Module Walnut Pi CM2 (Allwinner T527) Compute Module Critical Link MitySOM-AM62A SoM Avaota F2 RISC-V AI Camera SBC