Only one model fits on the graphics card at a time, but switching takes about 45 seconds and your chats stay as they are.
Switch from the chat screen
Click the model name at the top of the chat and choose the other model.
Compare them on the Models page
Models shows both models side by side: speed, how much faster they are than standard llama.cpp, and what each is best at. Use Switch to this model, Reload or Eject (frees the graphics card).
5090Ai
Choose where the model files are stored
The models take about 45 GB. To keep them on another drive, open Settings → Model files and click Change…. 5090Ai can move existing files for you.
On fast internet, more download connections can speed up downloads.