There is a very long thread on this topic with lots of slightly different code snippets and conflicting sets of step-by-step instructions. I believe I finally got it working, at least it looks like it, but I'm not sure if it's really helping performance. I have a test scene with a few static objects that use a baked light map. At runtime I instantiate about 25 different prefabs in the scene. Before I starting trying to bake the prefabs, the Stats window showed 79 batches, 57.3k tris, and 33 SetPass calls. It runs at about 240-250 fps in the Editor. That's my baseline. I took 6 of those prefabs that all share the same material, set them to static, and baked their light maps as discussed in the thread linked above. The scene uses 4 non-directional light maps, 64 MB. Aside from the baked prefabs being a bit dark, it appears to be working. However, now that test scene shows 1298 batches and 35 draw calls, and the frame rate is up slightly to around 250-260 fps. This is not a big sample, so I wasn't expecting big effects on performance, but I didn't expect batches to go up, let alone by 16x. And I didn't expect SetPass calls to go up either. I'm not sure the slight difference in frame rate is significant. I was expecting fewer batches and fewer SetPass calls. Isn't that what should happen? Does it sound like I'm doing something wrong with the prefab baking? I don't want to spend a lot of time tweaking the lighting and baking the rest of my prefabs (about 100 different objects) if it's hurting performance.