The Axelera Model Zoo in one place: convolutional and transformer vision models, and the vision-language and language models, all read from the published Model Zoo doc.
Every vision model in one sortable, filterable grid — task, licence, input size, accuracy, quantisation loss and throughput per card.
Vision · ParetoThe trade-off chart, scoped to one dataset at one input size so the axis is honest. Optional Pareto frontier.
VisionLatency and throughput per frame for the transformer vision models, on their reference platform.
Generative AITime to first token and decode rate, with each model's context and prompt length, on PCIe and M.2 reference platforms.