![]()
Sipeed is focus on RISC-V and AIoT/tinyML Deploy, provide highly cost-effective AIoT solutions.
Sipeed MAIX is Out-of-Box TinyML/AIoT Platform for everyone, composed of Hardware(M0,M1,M2,M3...),Software(TinyMaix,MaixPy,MaixHAL,MaixBox), Service(MaixHub model/app shop).
In this project, we will introduce new rich-featured Hardware M1s and M0sense:
M1s: 480M RV64(C906)+ 320M RV32(E907) + 160M RV32(E902) Heterogeneous Tri-core AIoT MCU, 64MByte UHS PSRAM(2000M clk), 100GOPS NPU, WIFI/BLE/Zigbee, USB OTG HS,MIPI Camera, RGB LCD, also with basic Linux support!
M0sense: New typical TinyML board with Auido/IMU sensors, support KeyWords Spotting/Human Activity Recognition/Gesture Recognition applications.
![]()
They are not only highly cost-effective hardware, but also Out-of-Box AIoT platform:
drag and drop apps, lightweight python script, and easy share model/app stores~
We also prepare efficient inference lib and online trainning service for AIoT/tinyML enthusiasts:
TinyMaix: Tiny inference Neural Network library specifically for MCUs(TinyML), 400lines core code, deadly easy to port, 40+ chips ported!
MaixHub: Online AI model trainning/sharing platform, support semi-automatic labeling, make your AI model trainning easier!
![]()
Ready for RISC-V AIoT/tinyML travels? Here we go~
(Note: See FAQ to change to slower but cheaper express)
![]()
We have M1(K210) several years ago, why we make new AIoT hardware?
K210 is the first RV64 AI chips, have many limits(RAM/OPS/Resolving/Peripheral);
ESP32-S3 is the typical IoT chips, but limited in CPU/AI/RAM.
M1s based on newest BL808, combine advantage of AI and IoT chips, but still keep the same price, it is the Sweet Spot of AIoT hardware!
![]()
![]()
We also make tiny M1s Dock (~1x2 inch) for most M1s functions demonstration!
![]()
![]()
![]()
Besides special NPU chips, we also awared make TinyML available for every MCU is important.
So we make TinyMaix inference library and typical RISC-V TinyML board M0sense.
M0sense have most tinyML peripherals, able to run typical tinyML models, but it smaller than 1 square inch, only priced at $4 !
It is the best gift for tinyML newbies~
![]()
![]()
![]()
![]()
More and more powerful, interesting chips coming out, have no time to learning new SDK development?
No problem, we provide several No-SDK way to play with new advanced chip~
1. Drag and Drop Apps
M1s Dock have USB2.0 HS, M0sense have USB2.0 FS, they can act as virtual mass storage to implement Drag&Drop Apps.
![]()
1. We
2. Python Script
Don't need any SDK, play functions with lightweight python script~
![]()
3. Support Dynamic C module
no need SDK, import&run simple C module in shell~
![]()
4. OpenSource SDK
Still open SDK for senior developers~ include AI toolchain~ (release in Nov.)
![]()
![]()
5. Linux Supported
M1s C906 core support standard Linux! cheers!
![]()
![]()
Try AI on the edge! We called it as AIoT(with NPU) or tinyML(without NPU)
We develop new tinyML inference library TinyMaix which suit for almost all MCUs~
TinyMaix is the most Easy-Porting tinyML library ever.
It is so easy to port and use, that we get more than 40 chip/platform supported in one month, most of them are community contribution!
- Core Code less than 400 lines, code .text section less than 4KB
- Low ram consume, even Arduino ATmega328 (32KB Flash, 2KB Ram) can run mnist with TinyMaix~
- Support INT8/FP32/FP16 model, experimentally support FP8, convert from keras h5 or tflite.
- Supoort multi architecture accelerate: ARM SIMD/NEON/MVEI,RV32P, RV64V, CSKYV2, X86 SSE2
- User-friendly interfaces, just load/run models~
- Support Full Static Memory config
![]()
We also provide Out-of-Box online model training/sharing platform: MaixHub
![]()
It support nearly 10 different inference platform, and provide rich models:
![]()
You can train your own classify/detection/kws model on it, even with semi-automatic labeling tools
![]()
Enjoy models from maixhub, depoly on M1s/M0sense:
M1s Dock Models
Simple MNIST
![]()
Mobilenet 1000 classes classification
![]()
Face detection
![]()
Face 68 keypoints
![]()
Hand Keypoint
![]()
Body Keypoints
![]()
M0sense Models
Key Word Spotting
![]()
Magic pen (Gesture)
![]()
Abnormal vibration Detection
![]()
![]()
![]()
![]()
![]()
![]()
![]()
![]()
![]()
![]()
![]()
Want this RGB LCD Convertor Board?
We will unlock it when crowdfunding amount exceeds 20K USD~
![]()
WIKI: wiki.sipeed.com
Forum: https://www.reddit.com/r/Sipeed
Telegram Group: https://t.me/sipeed
QQ group: 592731168
FAE support email: support@sipeed.com
![]()
Sipeed is an young team from Shenzhen, China.
We are expert at IoT Application, Embeded System, Deep Learning.
We aim to build full stack AIoT/tinyML platform, make AI easier for every one.
Let's Sipeed up, Maximize AI's power!