Time Complexity

Floyd-Warshall has three nested loops, each running $n$ times. The innermost loop body is $O(1)$. Total time: $O(n^3)$. This is slower than single-source Dijkstra ($O(n \log n + m)$), but faster than running Dijkstra $n$ times on dense graphs.

For sparse graphs with many queries, Floyd-Warshall is often simpler. The simplicity of the code makes it a practical choice when $n \leq 500$ or so. Beyond that, consider Johnson's algorithm.

Space complexity is $O(n^2)$ for the data structures used.