Introducing STM32CubeAI Studio

[Maxime_MARCHETTO1]

STM32Cube AI Studio is now available as ST’s new standalone environment for optimizing, validating, and deploying AI models on STM32 MCUs.  

Evolving from the X-CUBE-AI legacy, it brings a unified workflow, improved performance evaluation, and a modern user experience designed for both embedded engineers and AI developers.
 

What’s new for developers 

A stronger foundation for embedded AI 

• Built on the trusted X-CUBE-AI legacy to ensure consistent, reliable embedded AI performance 

• Based on ST Edge AI Core technology for aligned performance across the STM32 AI toolchain  

• Separates embedded and AI tasks to reduce project complexity  

• Modern, intuitive UI for smoother, faster development 
  
  

New capabilities for embedded model optimization, validation & acceleration 

• Detailed model performance reports, including latency and memory usage 

• Experiment comparison tools to understand design trade‑offs 

• Accuracy validation on both host and STM32 target devices 

• Layer-level inference metrics visualization for deeper optimization 

• Optimized model compilation for STM32 devices with the Neural‑ART Accelerator 
  
  

Smooth integration within the STM32Cube ecosystem 

• Fully compatible with STM32Cube tools 

• Reuses existing STM32 projects to simplify upgrades and transitions 

• Generates code and templates compliant with STM32 development guidelines 

• Provides secure unified authentication and API access for seamless integration with other STM32Cube tools 
  
  

Additional resources 

• Download it here

• Watch this tutorial and get started quickly 

• Read the wiki articles  

First published on Feb 18, 2026

[Tuomas95]

What if i want to use it with CMake and VSCode?

[Julian E.]

Hi @Tuomas95,

 

For now it is not supported.

I will contact the dev team to express the need.

 

Have a good day,

Julian

Nice to see ST moving toward a cleaner, more focused workflow for embedded AI. Separating the AI and firmware parts should really help reduce the usual debugging headaches, and the detailed latency/memory reports sound especially useful when squeezing models into tight MCU limits. The layer-level metrics and experiment comparison tools look like they’ll save a lot of trial-and-error time, too. Overall, it feels like a solid step forward from the old add-on approach.

[Julian E.]

Hi @unknown,

Thank you very much for your positive feedback!

We plan to publish more tutorials and gradually add new features over time.
In the meantime, please feel free to share any suggestions or feedback that could help us improve the tool.

Have a great day,
Julian

[UHjor.1]

No Linux version?

[Steven-LIN]

I would also like to know whether MPU will be supported in the future.

[L_Ricardo]

I downloaded the software expecting it to support the STM32N6570-DK board. Unfortunately, I couldn't find it to start a new project. Will this support be available in the future? If not, I'm having trouble deploying models based on yolov8n and yolov11n. I couldn't proceed with the tutorials or official guides or ST forums. I am having trouble with the screen not starting (black screen) after loading my model and other files, as instructed. Any suggestions or other posts that can help me with this? I haven't found anyone who has been successful with this yet.

[Tuomas95]

@L_Ricardo I do see the STM32N6570-DK when starting a new project.

[L_Ricardo]

Hi @Tuomas95 . Thank you for your reply... I tried it in "Development boards" and "MCU", but it return "No item found.". Maybe i'm checking in wrong place?

 

Edit: I also have the necessary programs, i guess... Maybe its because i dont have IAR Embedded  and Keil MDK-ARM? 

[Tuomas95]

@L_Ricardo 

I don't have IAR or Keil either, maybe try to reinstall?