ClikaRT: The inference runtime

Run your AI on every device you own.

One runtime for every one of them: the same model and the same code, whether it runs on a phone or on a rack of GPUs.

  • any AI model
  • any modality
  • CPU
  • GPU
  • NPU Coming
  • TPU Coming
  • Linux
  • Android
  • Windows
  • macOS
  • iOS Coming

The problem

Model × hardware × OS: the combinations explode.

Today each combination is a separate port that someone writes, tests and maintains, and the list grows with every new model and every new piece of hardware. ClikaRT replaces all of them with one runtime that adapts to the machine it runs on.

Our own engine

Nothing between your model and the hardware.

ClikaRT does not sit on top of a vendor's runtime; it is the runtime. So a new model or new hardware never waits on someone else's schedule, one install brings everything it needs, and changing hardware changes nothing in your code.

No setup

It finds your hardware and puts it to work.

Nothing to configure. ClikaRT reads what the machine has, the processor, the graphics hardware, the AI accelerator, and runs the model where it runs best, across several of them when one is not enough.

No conversion

Run the model file you already have.

No export step, no converted copy of the weights, no second format to maintain. The file you downloaded is the file that runs, on any machine you own, and the tokenizer, the sampling and the serving come with it.

Two ways in

As a server your apps call, or inside your app.

As a server, your apps call it over the network in the same request shape as the big AI services, so switching is an address change. Embedded, the model runs inside your app itself. Same model file, same answers, your choice per device.

Every generation

Old servers and new laptops run the same file.

ClikaRT ships one binary and decides at run time which instructions the hardware in front of it understands, from a server bought a decade ago to a laptop bought this year. Nothing to rebuild when the hardware changes.

The Private
Preview is Live

Leave your email and we will get you in.