Are you looking for actual bodies in the call or endpoints?
We often use the system to connect up our large conference/auditorium rooms across three main campus locations, the UK, Ireland and a number of our larger US field offices. Each main campus room holds from 1200 to 2500 people and there might be 100 or so in each other location. The bigger sites use Barco video walls to suit the crowds.
Generally, we will mute all endpoints, but there are inevitable Q&A sessions where they will round robin from site to site. We man these events and generally have the site with the current speaker in Lecture mode and change that as they go from site to site. It can get confusing and often smooth operations are dependent on the weakest link in the mix (such as one of the sites not having mics turned on, etc) but generally we've gotten things down to a (albiet imperfect) science.
So for a final number, our biggest yet consisted of about 6000 people- not counting the folks that watched it via webcast (work from homes)
Not really the standard conference call- but a viable nonetheless.
Michael O. Libby, CTS