Ollaya runs typed decision models on your own machine

Ollaya packages several open decision models behind a local server with TypeSafe-compatible request and response shapes. Applications can submit text or JSON and ask for choices or scores without generating a prose answer. It supports models including Laya and decoder-based classifiers, giving developers a way to compare local decision backends without rewriting every caller.

Ollaya

Why it matters

A substantive follow-up to Laya and Jev coverage: the new value is local packaging and API compatibility across models.

More context

Try the beta with a supported desktop, CLI or Docker installation and downloaded model weights. CPU execution is supported; GPU support varies by platform and model. Original launch date is unknown.

Keep in mind

Independent of both TypeSafe and Ollama. Compatible JSON does not establish equivalent accuracy or calibration. Published speed comparisons mix local GPUs and a hosted API; 1min has not reproduced them.

Original read