Inputting an image in the Assistant API using the new vision model ...

Responsive image
source