Specifications
Platform
NVIDIA Jetson AGX Orin 64GB module
AI Performance
275 TOPS (INT8, Ampere GPU + 2x NVDLA v2)
Memory
64GB LPDDR5
Onboard Storage
64GB eMMC, plus M.2 NVMe and MicroSD expansion
Camera Inputs
8x GMSL2 via two 4-pin Mini FAKRA connectors
Video Decode
Up to 22-channel 1080p, or 8K at 30fps H.265
High-Speed Networking
1x 10GbE RJ45
Additional LAN
5x Gigabit Ethernet RJ45
Wireless
Wi-Fi 6, Bluetooth 5.2
Cellular / GNSS
4G/5G cellular and GPS/GNSS support
I/O
USB 3.0, HDMI 2.0, RS-232, RS-485, CAN
Operating System
Ubuntu 22.04 with full desktop Linux environment
AI Frameworks
CUDA, TensorRT, cuDNN, PyTorch, TensorFlow, PaddlePaddle, MATLAB
LLM / VLM Support
DeepSeek-R1, Llama, ChatGLM, Qwen, Phi
Vision Models
EfficientViT, NanoSAM, TAM; Flux and Stable Diffusion XL image generation
Tooling
Ollama, ComfyUI, Docker containerisation, custom operator development
Pricing
$3,369 USD (single unit)
Stock
Available
Overview
The Firefly EC-AGXOrin is an NVIDIA Jetson AGX Orin 64GB-powered AI inference system built for edge deployments that need both high inference throughput and a large number of synchronised cameras. The platform delivers 275 TOPS through the AGX Orin module's Ampere GPU and dual NVDLA v2 accelerators, backed by 64GB of LPDDR5 RAM.
Its camera architecture is the standout capability: eight GMSL2 inputs delivered through two 4-pin Mini FAKRA interfaces, supporting up to 22-channel 1080p decoding or 8K at 30fps in H.265. Networking is equally generous, with one 10GbE RJ45 port plus five additional Gigabit ports, alongside Wi-Fi 6, Bluetooth 5.2, 4G/5G cellular and GPS/GNSS for mobile installations.
Industrial connectivity covers USB 3.0, HDMI 2.0, RS-232, RS-485 and CAN, making integration into vehicle and machine-control architectures straightforward. Software runs Ubuntu 22.04 with hardware-accelerated graphics and the full CUDA, TensorRT and cuDNN stack; Firefly documents support for ROS-based robotics, LLMs including DeepSeek-R1, Llama, ChatGLM, Qwen and Phi, vision models such as EfficientViT, NanoSAM and TAM, and image generation with Flux and Stable Diffusion XL.
At $3,369 per unit the EC-AGXOrin sits in the same class as the AAEON BOXER-8645AI and Vecow RAC-1000 rugged edge AI systems. QS Compute supplies volume and configuration-specific quotes.
Key Benefits
275 TOPS in a deployable edge chassis — AGX Orin 64GB performance without a rack. Eight GMSL2 inputs with Mini FAKRA deliver shielded, automotive-grade camera links over long cable runs. Six Ethernet ports including 10GbE support separate camera, control and backhaul network planes. Full NVIDIA software stack — CUDA, TensorRT, cuDNN, ROS and Ollama are all documented. Vehicle-ready I/O — CAN, RS-485, GPS/GNSS and 4G/5G suit mobile and in-vehicle programmes.
Applications
In-vehicle computing and autonomous driving development, robotic control and AMR fleets, 8-camera machine vision inspection, intelligent video analytics, mobile robotics with cellular backhaul, and local LLM/VLM inference for robotics and industrial assistants.
Request a Quote — FIREFLY EC-AGXORIN — NVIDIA JETSON AGX ORIN 64GB AI INFERENCE SYSTEM WITH 8X GMSL2
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →