About half my team is technically collocated in one location, there's a couple other locations with two people, and a handful of others. I say "technically collocated" because at any given time, half those people work from home anyway. This came about from an acquisition.
At first we did the meeting room plus remote thing. The best room we have for that has a ~60" tv and a Logitech Group camera/speaker/mic, and sits about 6-8 people. Plugs in with usb, and if the person plugging in does it right, you can get a decent view of all cameras and share a screen (trick is, the person plugging in can't be the one sharing). As a remote attendee the audio is good and the camera covers the room well.
What's happened over time though is we stopped using the room, and everyone joins from their desk (and people started working remote more). It puts everyone on a level playing field, and frankly, the meetings are better. I love being able to look things up or update a ticket on my second monitor with my full desk setup, and it's much less disruptive to the type of people who get offended if you're heads-down looking at your laptop while they're talking.
The only thing I really like better in person is standing around a whiteboard while designing or explaining something complex. Drawing on a touchscreen is the next closest thing I find works as a substitute but it's just not the same.
Tips:
Make it a requirement that every meeting invite gets an online session, even if everyone is in the same office. This helps shift things remote, if you don't or can't outright mandate the "one person remote means everyone is remote" rule, plus it avoids having to coordinate where anyone is.
Use video, and make it a habit to turn on your camera, as it prompts others to as well. Staying muted is good etiquette but makes it impossible to gauge reactions, so video lets you see people nodding, looking confused, laughing, etc.
Make sure everyone has a good audio setup (which basically means headset). For what it's worth, of non-headset users, maybe 20% have "passable but not great" audio, 2% have good audio (no feedback, echo, background noise), and the rest are garbage.