Developers and engineers who need fast AI processing at the edge often face a trade-off between power and size. A bulky desktop or a slow single-board computer can slow down robot vision, smart camera, or on-site data analysis, costing time and money. The Firefly AIBOX-OrinNX gives you a compact, high-performance platform that runs large models locally, removing the need for a cloud connection and keeping your projects responsive.
Firefly AIBOX-OrinNX Jetson Orin NX Embedded PC 16GB/128GB Standard without Wifi with Firefly
Meet the most powerful edge AI box you can mount on a robot arm or embed in a kiosk. It is built on NVIDIA Jetson Orin NX, delivering 157 TOPS of AI performance in a 93 mm square board. If you have struggled with slow inference or limited memory, this device solves the problem with 16 GB RAM and 128 GB storage, ready for demanding vision, language, and robotics workloads.
Key Features
- Octa-core Arm Cortex-A78AE CPU up to 2.0 GHz.
- 1024-core NVIDIA Ampere GPU with 32 Tensor Cores.
- 157 TOPS AI performance for demanding edge workloads.
- 16 GB LPDDR5 RAM and 128 GB eMMC storage.
- Supports ROS, LLaMA2, ChatGLM, Vision Transformers, Stable Diffusion.
- Docker container management for quick image deployment.
- HDMI 2.1 output - 4K at 60 fps for clear visual feedback.
- Power input 12 V DC, normal consumption 7.2 W, max 33.6 W.
- Two USB 3.0 ports, external watchdog support.
- Runs Ubuntu 22.04 with CUDA, TensorRT, cuDNN libraries pre-installed.
Specifications
| AI Performance |
157 TOPS |
| Container Management |
Docker container management technology for easy image deployment |
| AI Software Stack |
NVIDIA Jetson Orin provides high AI computing performance, large unified memory, and comprehensive software stacks.
Supports advanced generative AI applications, transformer-based models, and delivers high edge AI performance with
MLPerf-optimized inference
|
| Large Model Support |
Robotic Models: ROS models
Large Language Models: Private deployment of ultra-large-scale Transformer-based models such as LLaMA2, ChatGLM, and Qwen
Vision Models: ViT, Grounding DINO, and SAM
AI Painting: Stable Diffusion V1.5 image generation model
|
| CPU |
Octa-core 64-bit Arm Cortex-A78AE v8.2 processor, up to 2.0GHz |
| GPU |
1024-core NVIDIA Ampere architecture GPU with 32 Tensor Cores |
| Power Input |
DC 12V (DC 5.5 × 2.1mm) |
| Power Consumption |
Normal: 7.2W (12V/600mA)Maximum: 33.6W (12V/2800mA) |
| Video Output |
1 × HDMI 2.1 (4K @ 60fps) |
| Operating System |
Jetson system based on Ubuntu 22.04 with desktop Linux environment, accelerated graphics,
and support for NVIDIA CUDA, TensorRT, and cuDNN libraries
|
| Dimensions |
93.4 mm × 93.4 mm × 50 mm |
| Weight |
≈ 500g |
| Watchdog |
Supports external watchdog |
| USB |
2 × USB 3.0 (Maximum: 1A) |
