What is Mythic?
Analog compute-in-memory chips for efficient inference.
Who is Mythic best for?
Mythic is best for hardware teams building cameras, drones, robots and other edge devices that need AI inference at a few watts.
Edge device makers
The M1 chip runs 25 TOPS of inference at three to four watts and keeps up to 80 million model weights on the chip, so small devices need no external memory.
Video surveillance, drones and machine vision
The ME1 and MM1 M.2 cards add the same chip to existing boards, and the ME1 card measures 22 by 30 mm and draws under 5 W.
Automotive and ADAS
Mythic owns Videantis, a German AI processor maker whose silicon already runs in about 30 million production vehicles.
Teams with existing models
Its software takes models built in PyTorch, Caffe or TensorFlow and handles quantisation and compilation for the chip.
How much does Mythic cost?
Mythic does not publish prices: its M1 chip and M.2 cards are sold through a demo, an evaluation kit request or a product inquiry form, and the maker states no free plan or trial.
Mythic lists no prices for the M1 chip or the ME1 and MM1 M.2 cards; you book a demo, request an evaluation kit or fill in a product inquiry form that asks about your company, application and system requirements. Vanguard, Starlight and Mead are announced but not yet on sale. Plans and current prices are on mythic.ai.
How do readers rate Mythic?
No one has rated Mythic yet. Be the first.
Sign in to rate. One rating per account.
Questions about Mythic
What is Mythic?
Analog compute-in-memory chips for efficient inference.
Who is Mythic for?
Mythic is best for hardware teams building cameras, drones, robots and other edge devices that need AI inference at a few watts.
Is Mythic free?
No. Mythic is a paid tool.
Where can I try Mythic?
On its official website, mythic.ai.
Is Mythic software or hardware?
Mythic makes hardware: analog AI inference chips and M.2 cards that device makers build into their own products. It is not an app or an API you sign up for, although its software tools prepare PyTorch, Caffe and TensorFlow models to run on the chip.







