Supported Hardware¶
Bitweaver compiles and benchmarks models on a curated set of Arm Cortex-M class MCUs. Every part below has been validated end-to-end on real silicon in our benchmarking runners — uploads target these MCUs directly and return measured inference time, memory footprint, and flash.
More MCUs on the way
New MCUs are added continuously. If there is a specific MCU you would like to see supported, get in touch - new hardware is prioritized based on customer demand.
Supported architectures¶
Bitweaver targets Arm Cortex-M class MCUs. The following cores are currently tested and supported:
- Cortex-M33
- Cortex-M4
- Cortex-M7
Accelerator support¶
Where an MCU exposes a vendor AI/ML accelerator, Bitweaver can offload supported layers to it instead of running them on the Cortex core:
- Silicon Labs MVP (Matrix Vector Processor) - on the EFR32MG24 family
Supported vendors and MCUs¶
- MAX32655 familyCortex-M4
- CY8C624ABZI-S2D44Cortex-M4
- EFR32MG24 familyMVPCortex-M33
- STM32H7A3ZITxQCortex-M7
- STM32HL4R5ZITxCortex-M4
- STM32U5A5ZJTxQCortex-M33
Requesting new hardware¶
Need an MCU or vendor that isn't listed? New parts are added based on customer demand. Get in touch with the MCU part number and a short note on your use case, and we'll follow up on availability and timelines.