Rate limiting strategies, what's your default?

Hey everyone, I've been thinking a lot about rate limiting lately. APIs are everywhere now, and we really need to protect services from abuse or overload. There are tons of ways to do it, you know, like token bucket, leaky bucket, fixed window, and sliding window. For simple stuff, a fixed window counter works fine. But when things get complicated or you need more precise control, the other methods look better. What's your usual plan when you're designing a system that needs rate limiting? Do you have a favorite, or does it always come down to the specific needs?