The first box - the biggest (and arguably most important) box - is the social constellation. The real people of meat and bone and their arrangements and agreements. When and where do we meet, who brings the snacks, who's inviting whom, who is someone's boyfriend, that sort of stuff.
Once we get over that, the smaller box inside is the imaginary stuff we share: setting (the "world" of the game), characters (the people in it), situation (the particular arrangement and positioning of characters in the setting), system (for lack of a better word: the means by which we resolve or agree to what happens to these things - not "set of rules" as the word is usually understood in the hobby) and colour (the details with which we paint the above). At this point we also delegate authority, like who gets to say stuff about the setting or who comes up with which characters.
The even smaller box is how we actually do stuff, the procedures of play. Like "the GM prepares a map of the dungeon" or "you roll these dice and compare them to this number to see if your attacks hit" or "you get experience for doing discovering things man was not meant to know" and "when your character becomes level 7 she gets a castle".
The smallest box is the actual stuff that happens at the table, a more specific manifestation of the above. Rolling dice, looking up rules, talking in funny voices or talking in third person, the GM describing how the orc's head gets lopped off.
That's all simple right? We're not saying anything about dice pools vs roll under or sandboxes vs plot-based games or immersing in your character vs playing them like a pawn. That's all up to the individual group's arrangements, we're just putting the activity of roleplaying as observed in a set of four nested boxes.
Now, imagine an arrow piercing all the four boxes. It's the aim of play. It starts outside the social constellation of people sitting down at the table and ends up in the actual act of rolling dice or speaking in a funny voice. It's the motivation of "Let's play!" that starts with us arranging to sit down with certain people at a certain time and place and runs all the way down to what I'm rolling those dice for in the moment. If you imagine it like a skewer, you can also imagine how elements from the individual boxes get arranged on it. We pick out bits out of individual boxes (out of a billion options) based on what we want on our arrow.
You can probably imagine how certain arrangements work better than others, or how groups may favour some over others. If at the level of imaginary stuff we choose an arrangement of heroic colour, superheroes and so on, then a set of procedures where character failure is constant and catastrophic doesn't really support our initial setup (unless we're trying to do a comedy game about incompetent superheroes or something).
Note that the individual pieces we choose (like, choosing if we're playing in ancient Rome or far space, whether we're using a ruleset with dice pools or percentiles, whether players get to author stuff about the world or that's just the GM's privilege, all those things) don't necessarily say anything about our arrow. On the other hand, our arrow will certainly inform the choices we make all the way down.
And that's it, that's the Big Model in a nutshell. I think it's really simple. It's just a model of observation of the activity of roleplaying as a whole, no biggie. The part that trips most people up really hard is that the Big Model goes into detail about there being broad categories that the individual group/game/session's arrows fall into. That's the next post.