CORE-6490L QCS6490 AI System on Module

CORE-6490L QCS6490 AI System on Module

4G+64G
$349.00
Sale price  $349.00 Regular price 
Skip to product information
CORE-6490L QCS6490 AI System on Module

CORE-6490L QCS6490 AI System on Module

$349.00
Sale price  $349.00 Regular price 
Shipping calculated at checkout.
Spec:4G+64G

CORE-6490L is a QCS6490 AI SoM with an octa-core CPU, 12 TOPS AI performance, 4K video processing, and rich interfaces for robotics, machine vision, and embedded systems

  • American Express
  • Apple Pay
  • Google Pay
  • Mastercard
  • PayPal
  • Shop Pay
  • Union Pay
  • Visa
  • WeChat Pay
CORE-6490L System on Module

CORE-6490L System on Module

CORE-6490L is a Qualcomm QCS6490 SoM featuring an octa-core 64-bit CPU, 6 nm process, and up to 12 TOPS of AI performance. It supports 4K video processing and simultaneous input from five MIPI-CSI cameras. With PCIe 3.0, USB 3.0, enterprise-grade security, and Qualcomm Linux and Ubuntu compatibility, it is ideal for edge AI, machine vision, industrial control, surveillance, and robotics.

CORE-6490L
12 TOPS On-Device AI for Local Large Models

12 TOPS On-Device AI for Local Large Models

Powered by the Qualcomm QCS6490, the module integrates the sixth-generation Qualcomm AI Engine and Hexagon DSP, delivering up to 12 TOPS for image recognition, object detection, video analytics, and real-time inference. Hardware AI acceleration supports mainstream vision algorithms, private deployment of on-device large models, and generative AI—reducing cloud dependence and latency while enhancing data privacy at the edge.

Comprehensive Full-Stack Software Ecosystem

Comprehensive Full-Stack Software Ecosystem

Built on the Qualcomm Linux software stack with native Ubuntu compatibility, the platform supports private deployment of Transformer-based multimodal large models, including Gemma, Llama, Qwen, and Phi. It includes the QNN inference framework and supports TensorFlow, PyTorch, ONNX, Keras, and Caffe, enabling fast model conversion and validation for efficient end-to-end deployment of on-device AI applications.


Extensive AIoT Expansion Interfaces

Extensive AIoT Expansion Interfaces

Equipped with five MIPI-CSI interfaces, the module supports synchronized capture from up to five HD cameras. MIPI-DSI, DP, and eDP enable independent or mirrored dual displays, while PCIe 3.0, USB 3.0, and USB 2.0 provide high-speed expansion. QUPv3 offers up to 16 configurable UART, SPI, or I²C interfaces for industrial control and edge IoT devices.

Application Scenario

Widely applicable to edge computing, private deployment of on-device large models, robotics, industrial vision inspection, intelligent surveillance, in-vehicle multi-camera perception, smart city terminals, smart retail, self-service kiosks, industrial scanners, tablets, and other products and applications.

Edge Computing
Edge Computing
Private Deployment of Large Models
Private Deployment of Large Models
Robots
Robots
Smart IP Camera
Smart IP Camera
Self-Service Terminals
Self-Service Terminals
Industrial Scanner
Industrial Scanner

Interfaces

Interfaces

Specifications

CORE-6490L
Basic Specifications SOC

Qualcomm QCS6490

CPU

Qualcomm Kryo CPU: Octa-core 64-bit processor (1×Kryo Prime 2.7GHz + 3×Kryo Gold 2.4GHz + 4×Kryo Silver 1.9GHz), built on advanced 6nm process, max frequency up to 2.7GHz

GPU

Adreno 643 GPU, supports OpenGL ES 3.2, Vulkan 1.x, OpenCL 2.0, DX FL12

ISP

Spectra 570L ISP, Zero Shutter Lag (ZSL) mode: 64MP / 36MP+22MP / 3×22MP @30fps, Non-Zero Shutter Lag (non-ZSL) mode: 192MP

NPU

Hexagon 770 NPU, 12 Dense TOPS

Codec

Video Decoding: 1×4K@60fps/2×4K@30fps/4×1080P@60fps (H.264/AVC, VP9)

Video Encoding: 1×4K@30fps/4×1080P@60fps (H.264/AVC)

Video Concurrency: 1080P@60fps decoding + 1080P@60fps encoding / 4K@30fps decoding + 1080P@30fps encoding

RAM

4GB/8GB LPDDR5

Storage

64GB eMMC

OS

Qualcomm Linux, Ubuntu

Software support

Supports private deployment of ultra-large-parameter models under Transformer architecture, including large language models such as Deepseek-R1, Gemma, Llama, Qwen and Phi series.

Supports QNN AI inference framework and multiple deep-learning frameworks: TensorFlow, TensorFlow Lite, PyTorch, Keras, Caffe.

Power

DC 3.8V

Power consumption

Normal: 0.532W(3.8V/140mA), Max: 6.08W(3.8V/1600mA), Min(Sleep): 0.133W(3.8V/35mA)

Interface type

LGA package, total 463-Pin, 14-layer PCB design; Pad size: 1mm × 0.6mm; Module flatness control: 0.15mm

Dimension

43.0mm× 35.0mm× 2.33mm

Weight

≈7g

Environment

Operating Temperature: -20℃~60℃, Storage Temperature: -20℃~70℃, Operating Humidity: 10%~90%RH (No condensation)

Interface Specifications Internet

Ethernet, Wi-Fi / Bluetooth expandable via PCIe and USB interfaces

Video input

5 × 4-Lane MIPI-CSI, supports D-PHY v1.2 / C-PHY v1.2, enables simultaneous acquisition from 5 cameras

Video output

1 × MIPI-DSI (D-PHY 1.2 or C-PHY 1.2, 4-Lane, max resolution 4K@30Hz)
1 × USB DisplayPort 1.4 (max resolution 4K@30Hz)
1 × eDP 1.4 (max resolution 4K@30Hz)
Supports dual-screen different-display / dual-screen same-display

Audio
2 × SoundWire, 3 × DMIC, 5 × MI2S (each group with 2 data channels, supporting full-duplex stereo or up to 4-channel transmit-receive), 1 × MI2S (4 data channels, supporting up to 8-channel transmit-receive)
PCIe

1 × PCIe Gen3x1, 1 × PCIe Gen3 x2 (NVMe-capable)

USB

1 × USB3.0 Gen1 (5Gbps, DP video output supported), 1 × USB2.0

SDIO

2 × SDIO3.0

Other interfaces

16 × QUPv3 Bus (QUPv3 can be flexibly configured as UART, SPI or I2C. Signals are multiplexed and cannot be used simultaneously; allocate resources according to practical application)

Customization

Firefly team provides hardware, software, turnkey and OEM/ODM customization services.

Firefly is dedicated to delivering hardware-software-integrated intelligent solutions across diverse industries. Firefly is dedicated to delivering hardware-software-integrated intelligent solutions across diverse industries.

Contact Us

Onsite SupportOnsite Support

Email SupportEmail Support

Remote Video SupportRemote Video Support

Real-time SupportReal-time Support