Specifications
Board
Synaptics Coralboard (limited edition shown at Google I/O 2026)
System-on-Module
Grinn AstraSOM-261x
SoC
Synaptics Astra SL2619
CPU
2x Arm Cortex-A55 @ 2 GHz + 1x Arm Cortex-M52 @ 200 MHz
AI Accelerator
1 TOPS Synaptics Torq inference engine implementing a Google Coral NPU
System Memory
2 GB DDR4 @ 3200 Mbps (1 GB option)
Storage
MicroSD card slot
Camera
MIPI camera interface; Google I/O 2026 edition ships an Arducam OV5647 1080p mini camera module
Audio
I2S2 and I2S3 audio connectors (1.8 V)
Connectivity
Optional Wi-Fi and Bluetooth via M.2 module
Expansion 1
M.2 E-key or E+A key (SDIO 3.0, UART/PCM, GPIOs)
Expansion 2
mikroBUS connector; Qwiic expansion connector
USB
USB Type-C for power (5 V, no power negotiation) and data (device mode)
Debug
UART console for kernel messages, JTAG interface, THT holes for POGO-pin probe clip
Power Measurement
Shunt resistors and test points accessible by pogo pins from the add-on board
Battery
2-pin connector without charging functionality
I/O Voltage
3.3 V unless otherwise stated
OS
Yocto Linux distribution
SDK
Synaptics Astra SDK with bootloaders, kernel drivers and configuration; MLIR-based Synaptics Torq toolchain
Model Support
Runs Google Gemma 3 270M out of the box
Availability
Limited edition for Google I/O 2026; general availability and pricing to be announced later in 2026
Overview
The Synaptics Coralboard is a development board for the Synaptics Astra SL2619 Edge AI SoC, built on a Grinn AstraSOM-261x module. The SoC combines two Cortex-A55 cores at 2 GHz with a Cortex-M52 microcontroller core and a 1 TOPS Synaptics Torq inference engine that implements a Google Coral NPU.
It is a deliberately low-power, low-cost entry point into on-device language and vision models: 2 GB of DDR4, a microSD slot for storage, MIPI camera input, I2S audio connectors, mikroBUS and Qwiic expansion, and an M.2 E-key socket for Wi-Fi and Bluetooth. The board runs Yocto Linux and is supported by the Synaptics Astra SDK, and it executes Google's Gemma 3 270M model out of the box through the open-source, MLIR-based Torq toolchain.
The unit that shipped for Google I/O 2026 was a limited edition — the Google I/O edition adds an Arducam OV5647 1080p camera module and a mikroBUS sensor HAT with camera connector, microphones, a piezo buzzer and user LEDs. Synaptics has said general availability and pricing will be announced later in 2026, making this a board to design against rather than to source in volume today.
Key Benefits
Google Coral NPU via the Synaptics Torq engine — a direct path into the Coral software ecosystem. Runs Gemma 3 270M out of the box on a 1 TOPS, battery-class power budget. mikroBUS and Qwiic plus M.2 E-key for rapid sensor and connectivity prototyping. Power-measurement shunts and pogo-pin test points exposed for energy profiling.
Applications
On-device small-language-model prototyping, battery-powered vision sensing, Google Coral NPU application development, smart-sensor and smart-home device design, academic edge-AI research.
Request a Quote — SYNAPTICS CORALBOARD — ASTRA SL2619 EDGE AI DEV BOARD WITH GOOGLE CORAL NPU
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →