0
I regularly need to review long reports and pull out specific claims, caveats, and disagreements between sections. I’m considering Claude and ChatGPT for this, but a big context window alone doesn’t tell me which tool will be more dependable.
For people who have tested both on real documents, what matters most: whether the model can use the whole file, how accurately it cites passages, or how well it handles follow-up questions? I’m especially interested in ways to evaluate long-document analysis without relying on a quick demo or a single impressive answer.
