PrismML: Seed
Compact AI models designed to run efficiently on phones, robots and other edge devices.
The business
PrismML develops model compression techniques for running capable AI on constrained hardware. Its Bonsai 2 model uses ternary weights to reduce a 27-billion-parameter model to 5.9 GB; it retained 98% of the benchmark performance of the original. Smaller models can bring inference closer to phones, robots and other devices where power and memory are limited.