Alibaba XuanTie C950 – A powerful, RVA23-complaint 64-bit RISC-V core for Edge AI computing
Alibaba has introduced the XuanTie C950 high-performance, 64-bit multi-core CPU IP with an out-of-order superscalar microarchitecture, RVA23 profile compliant, and support for "all optional extensions" such as Vector Crypto, Zacas, and Zama16. The company also says the XuanTie C950 supports the proprietary XuanTie AME (Attached Matrix Extension) ISA and supports integration with the company's XuanTie TPE (Tensor Processing Engine) IP. The new 64-bit RISC-V core will be found in SoCs with up to eight cores per cluster, targeting high-performance applications, such as cloud computing, edge computing, and AI computing. XuanTie C950 specifications: Architecture - RVA23 Profile Up to 8x cores clocked at 3.2 GHz; 22+/GHz Specint2006 base, or a score of around 70 at 3.2 GHz Pipeline - Superscalar out-of-order microarchitecture with 8-wide decode Floating Point - RISC-V F/D Extension Vector - RISC-V Vector Extension v1.0 with Vector Crypto support Matrix - XuanTie TPE coprocessor integration (AME v0.5) Hypervisor -
LooperRobotics Insight 9 standalone spatial AI camera features D-Robotics RDK X5 SoC, supports ROS 2 (Crowdfunding)
LooperRobotics Insight 9 is an autonomous plug-and-play spatial AI camera designed for embodied intelligence, quadruped robots, and dynamic mobile platforms. Compared to typical USB depth cameras like Intel RealSense D435i or Luxonis OAK-D, which rely on a host PC for processing, the Insight 9 integrates a D-Robotics RDK X5 octa-core Cortex-A55 processor with a 10 TOPS AI accelerator, allowing it to run Visual SLAM (V-SLAM) and depth mapping entirely on-device. The camera features a "Tri-Eye Perception Matrix," which includes an 8.4MP Sony Starvis IMX415 RGB sensor with an ultra-wide 188° field of view, and two SmartSens SG0132 global shutter sensors for stereoscopic depth. Encased in a passively cooled CNC aluminum chassis, it is also equipped with an automotive-grade Bosch BMI088 IMU capable of 24g high-G tracking, making it suitable for the heavy vibrations of legged locomotion. LooperRobotics Insight 9 specifications: SoC – D-Robotics RDK X5 octa-core Arm Cortex-A55 processor @ 1.5 GHz;
Hugging Face’s Reachy Mini is an open-source AI robot for your computer or Raspberry Pi CM4
Better known for its artificial intelligence software solutions, Hugging Face unveiled the Reachy Mini open-source desktop robot last year. It is designed to deploy AI applications that interface with the physical world. The robot features a camera, four microphones, and a speaker, and can move its 6 DoF (degrees of freedom) head, rotate its body, or wave its antennas thanks to nine servo motors. Two versions are available: the Reachy Mini Lite designed for computers running Mac, Linux, and Windows, and the Reachy Mini Wireless autonomous robot, powered by a Raspberry Pi CM4, adding WiFi and Bluetooth connectivity, an accelerometer, and battery support. Both models share most of the same specifications: Reachy Mini's SDK can be found on GitHub. It's based on Python, but also supports JavaScript and Web apps, and can integrate with LLMs to easily build apps and publish them to Hugging Face. The SDK also features several
ESP-IDF v6.0 framework adds support for ESP32-C5 and ESP32-C61, preview for ESP32-H21 and ESP32-H4
Espressif Systems released the ESP-IDF v6.0 framework a few days ago with stable support for ESP32-C5 and ESP32-C61 SoCs, as well as preview support for ESP32-H21 and ESP32-H4 low-power wireless microcontrollers. The framework also implements a new ESP-IDF Installation Manager (EIM) to make the ESP-IDF installation easier, relies on the low-footprint Picolibc C library, adds security and tooling updates, as well as a few Wi-Fi enhancements, and the ability to update the bootloader over the air. Here are some of the ESP-IDF v6.0 highlights: ESP-IDF Installation Manager - Unified cross-platform tool to simplify the setup process for ESP-IDF and compatible IDEs. It's available as a graphical interface or a CLI for automation and CI/CD pipelines. You can check the installation instructions for your OS. Picolibc replaces Newlib for a smaller memory footprint and better performance on resource-constrained devices. Check the Newlib vs Picolibc comparison for details. Contrary to some of
CIX ClawCore Armv9.2 CPU family targets OpenClaw deployments
OpenClaw was just introduced a few months ago, but we've already seen several low-footprint implementations, and some companies even ship mini PCs preloaded with OpenClaw. But today, I was just informed that CIX had gone further, and introduced the ClawCore Armv9.2 CPU family specifically designed/optimized for OpenClaw. The family will be comprised of three main SKUs: ClawCore-P (勁螯芯 "Powerful Claw") - High-performance model with 12-core CPU @ 3.2GHz, Immortalis-G720 GPU, 45 TOPS AI compute, and support for up to 64GB LPDDR5 RAM. Aimed at high-parallelism, large-capacity scenarios. Shipping starts now in March 2026. ClawCore-A (智螯芯 "AI/Smart Claw") - Octa-core CPU @ 3.0GHz, 80 TOPS AI compute (expandable to 200 TOPS via PCIe AI card), up to 64GB LPDDR5. It's designed for 24/7 use, supports full-chain ECC, hardware security (encryption/key management), and enables up to 50% reduction in model token costs via local inference. In practise, 80 to 90% of requests
Radxa AICore DX-M1M M.2 2242 low-power AI module delivers 25 TOPS of edge AI performance for just 3W of power
Radxa AICore DX-M1M is a compact, low-power M.2 edge AI acceleration module built around the DeepX DX-M1M neural processing unit (NPU) and delivers up to 25 TOPS (INT8) of AI performance while consuming only 3W of power. Designed for industrial robot arms, autonomous mobile robots (ARM), edge servers, drones, and AIoT devices, the module delivers high-performance AI and ML capabilities without blowing the power budget. It relies on a PCIe Gen3 x2 interface and works with both x86 and Arm systems, including the Raspberry Pi 5 and Radxa ROCK SBCs. AICore DX-M1M specifications: AI Accelerator – DeepX DX-M1M neural processing unit (NPU) with up to 25 TOPS AI System Memory – 1GB LPDDR4X @ 4266 MT/s (on-chip, supports up to 8GB according to DeepX) Storage – 1Gbit QSPI NAND / NOR flash Host Interface - PCIe Gen 3.0 x4 (supports Gen 1/2/3 and x1/x2) via M.2 M + B Key connector
PycoClaw – A MicroPython-based OpenClaw implementation for ESP32 and other microcontrollers
PycoClaw is a MicroPython-based platform for running AI agents on ESP32 and other microcontrollers that brings OpenClaw workspace-compatible intelligence to resource-constrained embedded devices. We had already covered the C-based Miniclaw for ESP32-S3 SoCs, the PycoClaw's developer (Jonathan Peace) told CNX Software that it is a "full OpenClaw-compliant agent" that supports more LLM providers (OpenAI, Gemini, Ollama, etc.), interfaces with not only Telegram, but also ScriptO Studio and WebRTC, and offers features like OTA updates, extensions, and battery-optimized operation. The table below compares PycoClaw to OpenClaw, Nanobot, PicoClaw, NullClaw, and MimiClaw. MimiClaw still offers the lowest footprint and highest efficiency, but PycoClaw appears to offer many more features, including improved GPIO support. It works on ESP32-S3 with at least 8MB flash and PSRAM, ESP32-P4, and should soon support Raspberry Pi RP2350 boards with PSRAM as well. PycoClaw can be installed on supported hardware through a "one-click install" using a compatible web
High-end 25MP global shutter camera with 10GbE interface is designed for NVIDIA Holoscan platform
Leopard Imaging LI-IMX530-10GigE-NL is a high-end 25MP global shutter camera designed specifically for the NVIDIA Holoscan edge AI platform. The camera utilizes a 10GbE interface for high-bandwidth, low-latency data transmission, making it suitable for gesture recognition, iris scanning, head roll, and eye tracking. At the core of the camera module is the Sony IMX530, a 1.2-inch CMOS sensor with 5328 × 4608 resolution and a 2.74 μm pixel size. The sensor data is handled by a Lattice CertusPro-NX FPGA, and a Marvell 10GbE PHY takes care of high-bandwidth data transfer to GPU systems. The camera supports NVIDIA Jetson AGX Orin, IGX Orin, and Thor platforms. LI-IMX530-10GigE-NL specifications: FPGA – Lattice CertusPro-NX FPGA 52K to 96K logic cells 7.3 Mb total embedded memory External LPDDR4 memory support Up to 156x (18 x 18) multipliers within sysDSP blocks for AI/ML workloads 10 Gigabit Ethernet PCS blocks Image sensor Sony IMX530 Diagonal 19.3 mm
cnx-software.com
The New Yorker
