Akida™Development Tools
Building Edge AI takes hardware and software engineered to work as one.
BrainChip’s Akida development ecosystem pairs neuromorphic silicon and reference
hardware platforms — boards, boxes, and FPGAs — with the MetaTF software environment
that trains, converts, and deploys models directly onto that hardware — a complete,
integrated path, from chip to application, for building Edge AI solutions.
Build Edge AI with hardware and software designed to work together.
BrainChip’s Akida ecosystem combines neuromorphic silicon and development platforms with MetaTF software to train, convert, and deploy models—from chip to application.
Hardware Development Tools
Choose a platform by chip. Every board runs the same MetaTF workflow, so models move from evaluation to production without rework.
AKD1500 delivers sub-watt, always-on inference with on-chip learning. Start on a Raspberry Pi 5, an M.2 board or a PCIe card, then design the same chip into your product.
Raspberry Pi 5 Development Kit
Raspberry Pi 5 paired with the AKD1500 M.2
board. Build and test always-on edge AI with
on-chip learning on hardware you already know.
M.2 Card B+M Key
An M.2 accelerator built on AKD1500. It
integrates seamlessly with Raspberry Pi 5
and compatible host systems for ultra-low-
power AI inference at the edge.
PCIe Development Card
A PCIe card with the AKD1500 co-processor for on-device inference. Pairs with single-board computers and industrial PCs for prototyping.
A pre-configured platform for loading Akida IP configurations and neural models to evaluate performance and execution. System and chip designers can use it for demonstration,
emulation, validation and system integration.
Akida FPGA Development Platform
Load Akida IP configurations and models to evaluate performance. Supports CNN and TENN models.
PCIe Development Card
A compact PCIe board with the AKD1000 processor for prototyping, testing and deploying edge AI.
M.2 Development Card
Accelerates CNN models with event-based processing and runs standalone without a host CPU. Suited to wearables, sensors and IoT. Includes the Akida Runtime.
MetaTF Software
Development Tools
MetaTF gives developers one path from idea to deployment on Akida. Build, train, test and deploy neural networks with a Keras-inspired Python API, then run them in
the built-in Akida IP simulator or on Akida hardware, including AKD1500 boards and the Akida FPGA platforms. The same workflow carries a model from early
evaluation through final tuning to production.
The Four MetaTF Packages Contain
Model Zoo(akida-models)
Load quantized models, or build and train Akida-compatible models.
Quantization Tool(Quantize-models)
Quantize models to low-bitwidth weights and outputs.
Conversion Tool(cnn2snn)
Convert models to a binary format
that runs on Akida.
Akida Interface(akida)
Runtime, hardware abstraction
layer and software backend for
simulation and hardware.
Cloud Tools
Akida Cloud
A pre-configured remote environment for evaluating models on hosted Akida IP.
No hardware required.
