Apart from the signal plotting performance, one could mention the huge gap between guiqwt and matplotlib in terms of image visualization performance (e.g. showing 20,000 x 20,000 images and zooming in/out continuously in real-time with your mouse is possible) and features (contrast adjustment, cross sections, mask tools, etc.).
As for the interactivity, I'm not comparing it to Bokeh since the end-user should not be the same:
- As far as I know, Bokeh is for plotting data inside a web page.
- guiqwt produces Qt widgets that can be used to plot signals or show images à la MATLAB but it is especially suited for embedding such widgets in Qt desktop applications (to develop signal and image processing software).
So, to sum-up the main uses of this library, I'd say that it's mainly for developping desktop Qt-based signal and image processing software requiring fast visualization and interactive tools.