You can give it far more material in one go. Large decks, long reports, many files, or a big codebase can stay in one working context instead of being broken into smaller prompts.
It is potentially getting tasks done faster contributing to lesser token usage with much longer usable outputs – suited for full drafts, detailed analyses, long code generation, or document creation without cutting off too early.