See which models fit your GPU before you download

Eugene checks each model against this machine’s graphics memory (VRAM) and RAM before you download it.

Find a model that runs here

1. Open Discover. It opens on Works here now; the filter also offers Works with another engine and Everything.
2. Search, or paste a Hugging Face link. Each result’s dot says works here, other engine or not here.
3. Click the dot to see which engines can run it on this machine, each with its reason, and whether it fits, is tight, fits, slower or does not fit.
4. Download it, then press Run on its page.

Add a model hub

1. Open Settings → Models & downloads → Where to find models, and choose add a hub.
2. Enter the Hub address and its Access token. Each hub is sent only its own token.
3. Use the arrows to set the order, then Save.

Delete a model

1. On the model’s page, press Delete….
2. Read the list: every file that goes, and, under Kept, because other models use them, the files that stay.
3. Press Delete and free (with the size). It cannot be undone. A model a machine is running must be stopped there first.

Strata (experimental)

In Discover, choose Works with another engine, open a model from Strata’s list, and press Download and prepare for Strata. On a model’s page, Run may suggest Prepare for Strata, then run (recommended) when Strata suits this machine better. Strata runs text only, on one GPU, one request at a time, and its memory fit is unknown.