Specifications
Model
AIBOX-9075
SoC
Qualcomm Dragonwing IQ-9075
CPU
Octa-core Kryo Gen 6 (Cortex-A78C-based) application cores up to 2.36 GHz
Real-Time Cores
Quad-core Cortex-R52 up to 1.85 GHz
AI Performance
Up to 200 TOPS — Hexagon Tensor Processor (HTP) with quad HVX and dual HMX
GPU
Adreno 663 — up to 1.2 TFLOPS FP32, secure GPU compute; Vulkan 1.2, OpenGL ES 3.2, OpenCL 2.0 FP, Adreno NN Direct
VPU Decode
AV1 / HEVC / H.265 / H.264 / VP9 / MPEG-2 — 1x 8Kp60, 2x 8Kp30, 4x 4Kp60, 16x 1080p60
VPU Encode
H.264 / H.265 / HEIF / HEIC — 2x 4Kp60, 4x 4Kp30, 8x 1080p60; concurrent 2x 4Kp60 encode + decode
ISP
Qualcomm Spectra 690 (2x IFE + 5x IFE_L)
System Memory
36 GB LPDDR5 with ECC support
Onboard Storage
128 GB UFS 2.2
SSD Expansion
M.2 2280 (PCIe 4.0 x4) socket for NVMe SSD
Camera Inputs
8x GMSL2 via 2x 4-pin Mini FAKRA connectors
Video Output
HDMI 2.0 up to 4K @ 60 Hz
Networking
Dual 2.5GbE with TSN; optional Wi-Fi 6 + Bluetooth 5.2 via M.2 E-key; optional 4G/5G via mini PCIe + SIM
USB
2x USB 3.0 Type-A (max 1 A), 1x USB Type-C for debugging
Industrial I/O
36-pin 3.5 mm Phoenix terminal block: 2x RS485, 2x CAN-FD, 12x opto-isolated DI/DO
Antenna Mounts
Holes for 4x 5G/4G and 2x Wi-Fi antennas
Frameworks
TensorFlow, PyTorch, ONNX, Paddle, Caffe, DarkNet
Operating Temperature
-40 °C to 85 °C
Enclosure
Rugged aluminium
Target Workloads
Edge computing, private LLM deployment, robotics, computer vision
Overview
The Firefly AIBOX-9075 is an industrial edge AI box built on the Qualcomm Dragonwing IQ-9075 platform, delivering up to 200 TOPS through the Hexagon Tensor Processor. It pairs an octa-core Kryo Gen 6 application cluster with quad Cortex-R52 real-time cores, 36 GB of ECC-capable LPDDR5 and 128 GB of UFS 2.2 storage with an additional PCIe 4.0 NVMe socket.
The connectivity mix is unusually complete for an edge box: eight GMSL2 camera inputs over two Mini FAKRA connectors, dual 2.5GbE with TSN for deterministic networking, HDMI 2.0, optional Wi-Fi 6 / Bluetooth 5.2 and optional 4G/5G, plus a 36-pin Phoenix terminal block carcing two RS485 channels, two CAN-FD channels and twelve opto-isolated digital inputs and outputs.
The rugged aluminium chassis is rated from -40 °C to 85 °C, which places it in vehicle, outdoor and heavy-machinery environments rather than a climate-controlled cabinet. Firefly positions it for on-device inference — particularly private LLM deployment — robotics and computer vision, with TensorFlow, PyTorch and ONNX supported through the Qualcomm toolchain.
Key Benefits
200 TOPS in a sealed industrial box — Qualcomm HTP with quad HVX and dual HMX accelerators. 36 GB ECC LPDDR5 supports larger on-device language models than typical 8-16 GB edge systems. 8x GMSL2 camera inputs for multi-camera surround and machine-vision rigs. -40 to 85 °C with opto-isolated DI/DO, CAN-FD and RS485 for direct machine integration.
Applications
Private on-premise LLM inference, multi-camera machine vision, autonomous robotics, automotive and off-highway prototyping, industrial monitoring with isolated digital I/O and CAN-FD.
Request a Quote — FIREFLY AIBOX-9075 — QUALCOMM IQ-9075 EDGE AI BOX, 200 TOPS, 36 GB LPDDR5
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →