Could anybody tell me or point me to a thread that explains the difference of features between the MDL5 model file format and the MDL7 one? I've found that MDL7 supports bones and multi-texturing, but is there anything else that makes MDL7 so special? My vertex animated models get much bigger file sizes in the MDL7 format, so I'd like to know if MDL7 is really always worth using, or if I should use it just when I need multi-texturing and bones animation.

the main difference is that mdl5 packs vertex coordinates into 2 bytes and mdl7 uses floats which take up 4 bytes. mdl5 has less precision which can be noticeable if your model is quite big and has fine local details.

if you want to save hard disk space you can continue to use mdl5. i don't think there is any performance difference.

MDL5 format uses a 16-bit variable for each vertex position. This is decent for vertex-animated models because it saves a lot of disk space. MDL7 format uses a 32-bit floating point position for vertices which allows for an extreme physical size of the model. MDL7 format allows for bones animation whereas MDL5 doesn't. In most cases the MDL7 format is what you need. For smaller models that use vertex animation or small static things, MDL5 format is sufficient. Anything finer than 1/16 of a quant for positioning precision is otherwise meaningless. To be able to see that sort of detail to a reasonable extent, you need a 20° field of view with camera.clip_near set to 1.