Intelligence at every power level
Astra pairs NPUs and GPUs with scalable MPU and MCU compute, so you can fit on-device AI to the power and performance budget of your product.
Models for Real-World Applications
Access optimized AI models for vision, audio, language, and multimodal workloads running on Astra. Explore models, reference implementations, and development resources to accelerate application development and bring intelligent experiences to market faster.
Learn from Real-World Edge AI Projects
View development guides, application examples, and best practices from the Astra ecosystem. Discover how developers are optimizing AI models, deploying applications, and bringing innovative edge AI solutions to life.
Hands-On Learning for Astra Developers
Accelerate development with practical tutorials covering platform bring-up, AI workloads, application development, and deployment on Astra.
Why Build Edge AI on Astra
An AI architecture ready for what's next
Synaptics Torq with Coral NPU adds flexibility through software updates, so devices keep up with new models without new hardware.
Quantized and tuned for the edge
INT8, INT4, mixed precision and ultra-low 1.58-bit quantization, plus hardware-aware tools that map models onto the NPU and GPU, help you balance speed, accuracy and memory.
- FP32100% of FP32
- INT825% of FP32
- INT412.5% of FP32
- 1.58-bit4.9% of FP32
Start with a ready model, or bring your own
Pick from models already optimized for Astra, or compile your own PyTorch, ONNX, JAX or TFLite model with the Torq or SyNAP compiler.
