Here's an older study that includes Claude 3.5: https://www.databricks.com/blog/long-context-rag-capabilitie...?

ruler https://arxiv.org/abs/2404.06654

nolima https://arxiv.org/abs/2502.05167

They often publish "needle in a haystack" benchmarks that look very good, but my subjective experience with a large context is always bad. Maybe we need better benchmarks.