OrangePi AI Studio Pro

TRAINING AND INFERENCE IN ONE AI COMPUTE CARD

Deeply adapted for DeepSeek

352 TOPS of AI compute

Orange Pi and Huawei, building a new AI ecosystem together

Ascend AI technology route

“Dual-chip” power, up to 352 TOPS

96GB / 192GB LPDDR4X

Training and inference in one

Type-C (USB4)

Windows support coming

Handles a wide range of AI workloads

OrangePi AI Studio Pro follows the Ascend AI technology route, bringing the ARM core, AI core, Vector core and Image core together to provide general-purpose compute, very strong AI compute and a codec core — enough to cover both training and inference on one card. It carries 96GB or 192GB of LPDDR4X running at up to 4266Mbps.

The card brings together Type-C (USB4), a power connector, a power button, LED indicators and an RGB light strip on the chassis. It suits OCR, object recognition, face recognition, search and recommendation, multimodal large models, content moderation, VR, digital healthcare, IoT and many other AI applications.

It currently supports Ubuntu 22.04.5 with the Linux 5.15 kernel, and Windows support is on the way — answering the demand for accessible compute while lowering the cost of learning, development and operation.

Ascend AI technology route

Two Ascend 310 series processors, bringing the ARM core, AI core, Vector core and Image core together on one card.

“Dual-chip” power, up to 352 TOPS

Two chips are integrated on the card for up to 352 TOPS of compute, enough to take on a wide range of AI workloads without reaching for a server.

Runs on a large memory pool

96GB or 192GB of LPDDR4X at up to 4266Mbps, which covers the demands of heavy, memory-hungry applications and larger models.

Training and inference in one

It provides general-purpose compute alongside the AI compute, and carries the whole underlying software stack for both training and inference — so one card does both.

Windows support coming

Ubuntu 22.04.5 on the Linux 5.15 kernel is supported today, with Windows to follow. The aim is to put real compute within reach and lower the barrier to AIGC work.

Deeply adapted for DeepSeek

Supports local deployment of the DeepSeek-R1 distilled models. Running them offline at the edge gives efficient on-device intelligence while keeping data handling secure and under your control.

Typical applications for OrangePi AI Studio Pro

AI + work

An AI document assistant, meeting support, knowledge-base question answering and more — work gets done faster, the office experience improves, and complex tasks can be handled intelligently.

AI + learning

Manage large volumes of local files efficiently and accurately. Retrieval-augmented generation makes it suited to AI teaching and to local knowledge Q&A over your own documents.

AI + everyday life

A personal assistant that saves time, with large models behind efficient, high-quality visual work — AI image generation and editing, and a general-purpose life assistant.

Internet of Things

AI and IoT together: devices generate and collect large volumes of data across many dimensions, stored in the cloud and at the edge, then turned into insight through big-data analysis and higher forms of AI.

Intelligent transport

Congestion detection, accident detection, equipment fault detection and vehicle-side applications — improving situational awareness, decision-making and human-machine interaction, and making the data genuinely useful.

Product detail images

Notes

1. Product batches and supply change in real time, so the specification above may be revised. Please confirm against the actual product on the day of sale.

2. This page is translated from the manufacturer’s own product page. Availability and lead time in Pakistan are confirmed on enquiry.