top of page

DISCOVER PYRIX

The world's first one time programmable AI accelerator

PYRIX delivers next-generation AI acceleration with one-time programmable (OTP) neuromorphic arrays, optimized for binary neural networks (BNNs). 
 
Each PYRIX chip features a high-speed matrix-vector multiplication (MVM) engine, leveraging a 1.58-bit precision architecture with an integrated binary ReLU activation function for efficiency in low-power, high-performance AI applications.
​
​
 
PYRIX Series – Scalable AI Performance

PYRIX-32X
  • 1,024 Processing Elements
  • Ultra-compact, ideal for low-power embedded AI
  • Optimized for lightweight AI models in IoT, robotics, and sensor fusion

PYRIX-64X
  • 4,096 Processing Elements
  • Balanced power and performance for real-time inferenceIdeal for smart cameras, automation, and edge AI applications

PYRIX-128X
  • 16,384 Processing Elements
  • Designed for mid-range AI workloads in autonomous systems
  • Optimized for AI-driven control systems, robotics, and industrial automation 

PYRIX-256X
  • 65,536 Processing Elements
  • High-performance AI acceleration for large-scale applications
  • Ideal for autonomous navigation, defense AI, and advanced inference
 
​
PYRIX is the ultimate AI acceleration solution, redefining efficiency, scalability, and computational power for the future of AI hardware.
bottom of page