MYTHIC M1The chip that started a new chapter in computing.
Mythic's first Analog Processing Unit (APU) delivers 25 TOPS of AI inference at a fraction of the power of conventional chips. It is the leading analog computing silicon in customers' hands today, proven across automotive, robotics, defense, and edge AI applications.
FORM FACTORSOne chip deployed in multiple ways.
M1 CHIPThe standalone chip runs at three to four watts—as low as one to two watts for lighter workloads. Built for space and power-constrained edge environments.
ME1 M.2 A+E KEY CARDA compact 22mm x 30mm card delivering 25 TOPS at under 5W. Designed for embedded edge AI applications where space is at a premium.
MM1 M.2 Key CardThe same performance in a slightly larger M.2 form factor, ideal for video surveillance, industrial machine vision, drones, and edge servers.
the Mythic difference25 TOPS. 3 watts. No external memory required.
Mythic M1 stores up to 80 million neural network weight parameters directly on-chip and performs all matrix multiplication in place, eliminating moving data to external memory. The result is desktop-grade AI compute performance at a fraction of the power, consuming 1/10 that of a typical embedded AI processor and 1/100 that of a GPU.
SOFTWAREWorks with the frameworks your team already uses.
Mythic M1 supports models built in PyTorch, Caffe, and TensorFlow. Mythic's software workflow handles optimization, quantization, and compilation, so what works in development works exactly the same way in production.
See it in actionObject detection on Mythic M1 silicon.
Ready to build on the right architecture?
Evaluate the Mythic M1 today.