Query complexity measures the information an algorithm must obtain through queries to an oracle, such as labels, comparisons, gradients, or subset counts. Researchers design adaptive and non-adaptive algorithms, randomized sampling methods, and learning procedures, then prove lower bounds that show when fewer queries cannot suffice. These ideas apply to learning and clustering, convex optimization and sampling, combinatorial problems, and social choice. A central tension is how query needs change with the access model, rounds, memory, and whether guarantees are worst-case or distribution-specific.