playground
Two local models, one prompt, side by side — all on your GPU. Models are GB-scale, so the columns run one after the other.
system prompt & temperature
Because each model is gigabytes, only one can be resident on the GPU at a time — so the two
columns run sequentially (column A finishes, then B loads and runs). A fair way to feel the
quality/speed trade-off between, say, a 1B and a 3B model. Nothing leaves your device.