Local Runtime
A local runtime is the app that actually runs models on your machine — it loads a GGUF file, drives your GPU, and gives you a chat window or an API.
Between "this model fits your card" and a working chat sits one piece of software. The runtime loads the weights, manages memory and the KV cache, and exposes the model — as a desktop chat window, a background service other apps can call, or both.
The well-known options are closer siblings than competitors: polished one-click apps, terminal-first services, the bare engine for tinkerers, and web interfaces that sit on top. Almost all of them load the same GGUF files and lean on the same underlying engine family, so the choice is about interface taste and workflow, not about which models you can run.
Practically: pick one, download a model that fits your memory, and you are chatting locally in minutes — no account, no cloud, no per-token bill.
Where you see it on the radar
We keep a short, curated list of runtimes — what each one is, who it suits, and where to get it — instead of a directory of forty lookalikes.
Pick your runtime →Related terms