Microkernel
A kernel design where as little as possible runs privileged — drivers and filesystems move out into ordinary processes.
The kernel handles only the unavoidable: address spaces, thread scheduling and message passing. Everything else becomes a user-space service. The gain is isolation — a crashed driver takes down a process rather than the whole machine, and it can be restarted.
The price is that what used to be a function call inside the kernel is now messages between processes, with context switches attached. Performance problems dogged early attempts such as Mach, and the famous 1992 Tanenbaum-Torvalds debate was about exactly this. More modern designs such as L4 and seL4 have moved a long way from that picture, and microkernels dominate where safety matters most — QNX in vehicles, seL4 in avionics, and Apple's XNU carries Mach heritage.