Axon driver

The Axon driver gives you direct access to the Axon NPU without the abstraction layer of the nRF Edge AI Library API. Both workflows below use the same driver initialization and hardware, but they target different workloads.

Workflow	Training toolchain	Best for
Axon driver inference	Axon NPU TFLite compiler	Running compiled TensorFlow Lite models with custom inference pipelines and direct NPU control
Axon driver DSP intrinsics	None	Hardware-accelerated signal processing with pre-defined DSP functions, without a compiled model

Axon NPU library is included as part of the Axon NPU: Hardware Accelerated Machine Learning and is currently available on the nRF54LM20B device.

Subpages:

• Axon driver inference
  • Hardware requirements
  • Software requirements
  • Model compilation
  • Application development
  • Next steps
• Axon driver DSP intrinsics
  • Hardware requirements
  • Software requirements
  • Application development
  • Next steps