Specifications
NPU
NXP Ara240 discrete neural processing unit (DNPU)
AI Performance
Up to 40 eTOPS (equivalent TOPS)
Precision
INT4, INT8 and mixed-precision inference
On-module Memory
8 GB or 16 GB LPDDR4 (stores all user models)
Host Interface
PCIe with DMA support on the TechNexion Telos platform
Form Factor
Enclosed module, 75 × 60 × 25 mm
Power Consumption
12 W (typical)
Max Model Size
Up to 30B parameters (INT4)
Supported Models
YOLOv8, LLaMA 2, Stable Diffusion
ResNet50 Latency
2 ms end-to-end inference
Model Architectures
CNN, Transformer, LLM, VLM, VLA
Operating System
Linux runtime (NXP Ara SDK)
Frameworks
TensorFlow, PyTorch, ONNX
Security
Secure boot and root-of-trust processor
Design Support
TechNexion Telos platform with DMA
Vendor
TechNexion Ltd — NXP Gold Partner
Overview
The TechNexion TELOS-AI4000 is a compact, enclosed edge-AI accelerator module built around the NXP Ara240 discrete neural processing unit (DNPU). It delivers up to 40 equivalent TOPS of AI inference for vision, generative AI and other embedded workloads while drawing only about 12 W, making it suitable for fanless industrial designs.
The module pairs 8 GB or 16 GB of on-module LPDDR4 with a PCIe host interface and DMA support on the TechNexion Telos platform, so models and data stay resident on the accelerator rather than competing with host memory. TechNexion reports that the TELOS-AI4000 runs YOLOv8, LLaMA 2 and Stable Diffusion, with a ResNet50 inference latency of 2 ms.
Measuring 75 × 60 × 25 mm, the TELOS-AI4000 is a larger, fully enclosed form factor than the bare M.2 Ara240 modules from other vendors, targeting system integrators who need a ruggedised, drop-in accelerator for vision and generative-AI appliances.
Key Benefits
40 eTOPS of edge inference from the NXP Ara240 DNPU; 16 GB on-module LPDDR4 keeps up to 30B-parameter models resident; 12 W typical power suits fanless industrial enclosures; and a PCIe host interface with DMA simplifies integration into existing edge platforms.
Applications
Vision inspection and machine vision, generative AI and LLM inference at the edge, video analytics, robotics and autonomous machines, smart retail and industrial IoT gateways.
Request a Quote — TECHNEXION TELOS-AI4000 — 40 TOPS NXP ARA240 EDGE AI MODULE
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →