Like a hardwired synthesiser, it's a single indivisible synthesiser model. It most likely has a default signal path, so you probably don't have to patch anything together to make a decent sound.
Like a modular synthesiser, any default connections can be overridden using patch points. These allow various audio and control signals to be rerouted. Unlike a modular synthesiser, these patch points may be physically grouped together into a handy patchbay, or even a pin matrix. The name "semi-modular" a bit of a misnomer — it's not really modular at all, so much as patchable, the same way a modular synthesiser is patchable.
Some semi-modular synthesisers are released as a series of related devices that can be bought separately, but even these have a main base unit that works as a fully functional synthesiser on its own (such as the MS-20 and the Model-101), with an optional expander model (such as the corresponding MS-50 and the Model-102) adding bonus functionality of more niche use. Other semi-modular synthesisers (such as the VCS3) exist only as a single, self-contained unit, with no expander available.
Sequencer models are also a common add-on (such as the corresponding SQ-10 and Model-104), but even some hardwired synthesisers can be connected to external sequencers. What separates hardwired and semi-modular synthesisers isn't the ability to connect something at the start or end of a signal path, but rather in the middle. It's the patch points.
The more interesting, the more manipulatable a synthesiser is — and a [semi-modular] system like our System-100 is totally manipulatable — the less preset it is, if you like, then the more likely you are to stumble across something that actually appeals to you.
— Martyn Ware, Heaven 17, 1982
- "Penthouse and Statement" Tony Bacon, One Two Testing, Nov 1982, pp. 12—14