EN RU ES FR DE FA
SpacemiT K1 Application
2026-08-1310 min read

myCobot 280 RISC-V: World's First RISC-V Full-Stack Open-Source Robot Arm

Elephant Robotics and SpacemiT co-launched the 280 RISC-V — a 6-axis collaborative robot arm powered by the K1 RISC-V AI CPU. Full-stack open source from instruction set to robot control code, with on-device LLM inference and ROS2 support.

SpacemiT K1myCobotRISC-VRoboticsOpen SourceROS2Edge AI

Overview

In August 2026, Elephant Robotics (a Shenzhen-based collaborative robot specialist) and SpacemiT (a RISC-V AI CPU startup from Hangzhou) jointly announced the myCobot 280 RISC-V — billed as the world's first 6-axis collaborative robot arm built on a fully open-source RISC-V stack. The arm replaces the traditional ARM-based controller with SpacemiT's K1 chip, bringing RISC-V from the MCU and SoC world into the robotics application layer.

The collaboration is significant because it demonstrates a complete vertical integration: the RISC-V instruction set, the chip SDK, and the robot control code are all open. This gives developers unprecedented ability to inspect, modify, and extend every layer of the robot's software stack — something that was previously impossible with proprietary ARM-based robot controllers.

SpacemiT K1: The Brain Inside

The K1 is SpacemiT's first-generation RISC-V AI CPU, featuring 8 X60 high-performance computing cores. It is designed to bridge general-purpose CPU computing with AI acceleration, making it well-suited for robotics workloads that require both real-time control and on-device AI inference.

K1 Key Specifications

ParameterValue
ArchitectureRISC-V 64-bit (RV64GCV)
CPU Cores8x X60 high-performance cores
Clock Frequency1.6 GHz
AI ComputingSupports 0.5B-1B parameter LLM inference
LLM Inference Speed>10 tokens/s (on-device)
Process Node22nm
OS SupportBianbu Linux (based on Debian/Ubuntu)
GraphicsPowerVR (Imagination) GPU

The K1's ability to run LLMs locally is a key differentiator. Traditional robot arms rely on cloud-based AI services for vision and decision-making, introducing latency and privacy concerns. With the K1, the 280 RISC-V can perform visual tracking, voice interaction, and autonomous decision-making entirely on-device.

Robot Arm Hardware

The 280 RISC-V inherits the myCobot 280 series' compact desktop form factor while upgrading the compute platform:

ParameterValue
Type6-axis collaborative robot arm
Reach280 mm
Payload250 g
Weight860 g
End Effector MountLEGO + M4 compatible
AttachmentsSuction pump, gripper, vision module
ControllerSpacemiT K1 RISC-V AI CPU

The arm's lightweight design (860 g) and small footprint make it ideal for desktop use in classrooms, labs, and maker spaces. The end effector supports both LEGO studs and M4 threaded holes, enabling rapid prototyping with off-the-shelf components.

Full-Stack Open Source

The defining feature of the 280 RISC-V is its commitment to open source at every layer:

Open-Source Stack Layers

LayerOpen-Source Component
Instruction SetRISC-V ISA (open standard, royalty-free)
Chip SDKSpacemiT K1 SDK (publicly available)
Robot ControlElephant Robotics control code (open)
OSBianbu Linux (open-source, Debian-based)
Package Repoapt repository with pre-installed AI models
SimulationROS2 + RVIZ + simulation environments

Pre-Installed AI Models

The device ships with an apt repository containing pre-configured models:

Development Experience

The 280 RISC-V supports multiple programming paradigms to accommodate developers at different skill levels:

Programming Modes

ModeDescription
Drag-and-dropVisual block programming for beginners
Script programmingPython/C++ scripts for intermediate users
ROS2 integrationFull ROS2 + RVIZ for robotics engineers
SimulationSimulate before deploying to hardware

Out of the box, the arm comes pre-loaded with example content and libraries. New developers can complete a basic pick-and-place with vision recognition within minutes, while advanced users can leverage ROS2 topics, services, and actions for complex multi-robot coordination.

Quick Start: ROS2 Pick-and-Place

# Source ROS2 workspace
source /opt/ros/humble/setup.bash
source ~/mycobot_ws/install/setup.bash

# Launch the 280 RISC-V driver
ros2 launch mycobot_280riscv mycobot_280.launch.py

# List available topics
ros2 topic list
# /joint_states
# /gripper_controller/command
# /camera/image_raw

# Move to home position
ros2 service call /set_home std_srvs/srv/Trigger

# Pick-and-place with vision
ros2 run mycobot_vision pick_and_place.py \
  --object red_block \
  --target_zone zone_b

Application Scenarios

The 280 RISC-V targets several key application areas where its combination of open-source flexibility, on-device AI, and compact form factor creates unique value:

ScenarioHow RISC-V + AI Helps
Smart retailOn-device visual tracking and autonomous decision-making for demo kiosks
Interactive exhibitionsVoice interaction + LLM-powered responses without cloud dependency
EducationFull-stack open source for teaching RISC-V, robotics, and AI together
Industrial automationCustomizable control loops and deterministic RISC-V real-time performance
Maker/prototypingLEGO-compatible end effector + open SDK for rapid iteration

Why RISC-V for Robotics?

The 280 RISC-V represents a broader trend: robotics is emerging as a key application domain for RISC-V beyond MCUs and edge devices. Several factors make RISC-V attractive for robot controllers:

SpacemiT K1 vs K3: Robot Arm Upgrade Path

While the 280 RISC-V currently uses the K1, SpacemiT's K3 chip (released January 2026) offers a significant upgrade path for future robot arm iterations:

ParameterK1 (Current)K3 (Future)
CPU Cores8x X60 @ 1.6 GHz8x X100 + 2x real-time cores
AI ComputeSupports 0.5-1B LLMSupports 30-80B LLM (30x improvement)
StandardRV64GCVRVA23-compliant
Robotics BenefitBasic vision + voiceFull VLA (Vision-Language-Action) models on-device

The K3's dual real-time CPU cores can independently run a robot's ROS system while the main cores handle AI inference — a architecture that directly addresses the latency-critical control loop requirements of multi-joint robots. This makes the K3 a natural successor for a future "280 RISC-V Pro" or larger-payload robot arms.

Official Resources

Multi-Language Summary

EN: The myCobot 280 RISC-V is the world's first 6-axis collaborative robot arm with a fully open-source RISC-V stack, powered by SpacemiT K1 (8-core X60, 1.6 GHz). Supports on-device LLM inference (0.5-1B params, 10+ tokens/s), ROS2, and drag-and-drop programming.

RU: myCobot 280 RISC-V — первый в мире шестикоординатный协作ративный робот-манипулятор с полностью открытым RISC-V стеком, на базе SpacemiT K1 (8 ядер X60, 1.6 ГГц). Поддерживает локальный вывод LLM (0.5-1B параметров), ROS2 и блочное программирование.

ES: myCobot 280 RISC-V es el primer brazo robot colaborativo de 6 ejes del mundo con una pila RISC-V completamente de código abierto, impulsado por SpacemiT K1 (8 núcleos X60, 1.6 GHz). Soporta inferencia LLM en dispositivo, ROS2 y programación visual.

FR: myCobot 280 RISC-V est le premier bras robot collaboratif 6 axes au monde avec une pile RISC-V entièrement open source, propulsé par SpacemiT K1 (8 cœurs X60, 1.6 GHz). Il prend en charge l'inférence LLM sur appareil, ROS2 et la programmation visuelle.

DE: myCobot 280 RISC-V ist der erste kollaborative 6-Achsen-Roboterarm der Welt mit einem vollständig quelloffenen RISC-V-Stack, angetrieben vom SpacemiT K1 (8-Kern X60, 1.6 GHz). Unterstützt On-Device LLM-Inferenz, ROS2 und visuelle Programmierung.

FA: myCobot 280 RISC-V اولین بازوی ربات همکار ۶ محوره جهان با پشته کاملاً متن‌باز RISC-V است که توسط SpacemiT K1 (هشت هسته X60، ۱.۶ گیگاهرتز) قدرت می‌گیرد. از استنتاج LLM روی دستگاه، ROS2 و برنامه‌نویسی بصری پشتیبانی می‌کند.

Conclusion

The myCobot 280 RISC-V is more than a product launch — it is a proof point that RISC-V has matured enough to power real-world robotics applications. By pairing SpacemiT's K1 AI CPU with Elephant Robotics' proven mechanical platform, the collaboration demonstrates that the open RISC-V ecosystem can deliver competitive performance for latency-sensitive, AI-infused robotics workloads. For developers, educators, and makers, the full-stack open-source approach removes the last barrier to understanding and customizing every layer of a robot arm — from the instruction set up.