ADLINK DLAP-701

Specifications

Overview

The ADLINK DLAP-701 is a compact edge AI platform powered by NVIDIA Jetson Thor T5000 or T4000 modules. With up to 2,070 FP4 TFLOPS from its Blackwell GPU, 128GB LPDDR5X memory at 273 GB/s, and QSFP28 networking delivering 4×25GbE, the DLAP-701 brings data-center-class AI inference to space-constrained edge environments. Purpose-built for humanoid robots, autonomous mobile robots (AMR), and vision sensing systems (VSS), it integrates 4× CAN-FD interfaces, rich USB 3.2 connectivity, and industrial-temperature operation from -20°C to 60°C.

Announced March 2026 at GTC, the DLAP-701 marks ADLINK's first Jetson Thor platform — expanding their DLAP series from Orin-based DLAP-211 into Blackwell-era edge computing. The platform supports dual HDMI 4K output, M.2 expansion for Wi-Fi 6 and 5G, and a 128GB industrial-grade SSD with NVMe expansion via M.2 M-key. The 20-pin Euro terminal block provides isolated CAN-FD, UART, I²C, and auxiliary power rails for sensor and actuator integration in robotics applications.

Platform Variants

DLAP-701-T5000: Jetson Thor T5000 module, 2,070 FP4 TFLOPS, 128GB LPDDR5X, 2560-core Blackwell GPU with MIG support (10 TPCs). Optimized for multi-modal AI workloads — simultaneously running VLMs, LLMs, and world foundation models for complex humanoid robot perception and planning.

DLAP-701-T4000: Jetson Thor T4000 module, 1,200 FP4 TFLOPS, 64GB LPDDR5X, 1536-core Blackwell GPU with MIG support (6 TPCs). Cost-optimized variant for AMR navigation, industrial vision inspection, and single-workload edge AI deployments.

DLAP-711 (Sibling Platform): Same Thor T5000/T4000 base, differentiated by 4× GbE + 2× 100 Mbps Ethernet ports, smaller form factor, and extended operating temperature to -20°C to 65°C. Targeted specifically at humanoid robot applications with multi-network segmentation requirements.

Key Benefits

Blackwell Architecture: 5th-generation Tensor Cores with FP4 support and Multi-Instance GPU (MIG) deliver breakthrough inference for LLMs, VLMs, and multi-modal models directly at the edge — no cloud round-trip required. Quad CAN-FD: One galvanically isolated CAN-FD (3 kV) plus three with transceivers provide redundant, noise-immune communication for robotic joint controllers, motor drives, and vehicle networks. QSFP28 Fabric: 4 independent 25GbE MACs enable high-bandwidth sensor fusion across multiple GMSL2 camera streams, LiDAR, and radar pipelines. Industrial Compact: 211×194×94 mm footprint with wide 9-36V DC input range and -20°C to 60°C fanless operation for factory floor, warehouse, and outdoor deployment.

Applications

Humanoid Robots: Multi-modal AI perception (VLM + VLA), whole-body motion planning, real-time sensor fusion for dexterous manipulation. Autonomous Mobile Robots (AMR): SLAM navigation, obstacle avoidance, multi-camera vision pipelines, fleet coordination over 25GbE. Vision Sensing Systems (VSS): Industrial quality inspection, multi-camera AI analytics, real-time defect detection with sub-millisecond inference latency. Medical Imaging: On-device AI-assisted diagnostics, surgical robot perception, portable ultrasound/CT inference.

Related Products

ADLINK DLAP-211 Orin — Previous-generation Orin NX platform, 157 TOPS. AAEON BOXER-8741AI — Thor T5000 industrial box PC with 100GbE. Connect Tech Anvil-T5 — Thor T5000 multi-vision robotics platform.