Compress once, Run everywhere.

The inference runtime: a single API spanning CPU and accelerator backends, with tensors, operators, quantized weights, model loading and serving. C++ at the core, with Python and Kotlin bindings.

The command line interface: infer, serve or benchmark popular models, one command each, built on top of Modelverse, the CLIKA model library.

The CLIKA web platform: projects, models, benchmarks, deployments and licenses.