When should Python code use threads, processes, or asyncio?

Threads often suit blocking I/O when libraries expose synchronous APIs. asyncio suits large numbers of cooperating I/O operations when the libraries are asynchronous; blocking work still blocks the event loop unless moved elsewhere. Processes can use multiple CPU cores for CPU-bound Python work, with serialization and startup costs. CPython’s traditional GIL limits simultaneous Python bytecode execution in ordinary GIL-enabled builds, but extension modules and free-threaded builds change that picture. Base the choice on the actual interpreter, libraries, workload, and measurements rather than a universal rule.