Long documents
Working with very long inputs, such as long conversations and long stories. Good means the model keeps track of the right information across the whole input.
Sources last checked 2026-10-23.
We can't show scores for this kind of work yet. These testers publish results: Context Arena (Dillon Uzar) (one test) and Fiction.live (one test). See who tests it
At a glance
Vendor claims have not been collected for any category yet.
ⓘ More about this category
- Reliability not yet measured
- 2 sources, linked results only
- Sources last checked 2026-10-28
- Ring 4: needs rings 1 to 3 first
- MRCR = Context Arena (MRCR), FLB = Fiction.LiveBench
Tested by
Context Arena (MRCR)
Context Arena (Dillon Uzar), checked 2026-10-28A long chat holds several near-identical requests; the model must find and repeat the right one, in chats from a few pages to a few thousand pages long.
Last updated: not published by the tester. We last saw current models listed on 2026-10-28.Covers 21 of the models we trackIts results table includes a version of this task from Google DeepMind.Scores not shown: permission requested on 2026-10-27
See their resultsMore about this test: Context Arena (MRCR)
Reliability not measured by this source.
Technical details for Context Arena (MRCR)
AUC to 1M tokens (8 needles): A long chat holds several near-identical requests; the model must find and repeat the right one, in chats from a few pages to a few thousand pages long.
License: Score reuse not stated. Checked 2026-10-28.
Fiction.LiveBench
Fiction.live, checked 2026-10-28Questions about long stories that need real understanding, such as who knows a secret or how a relationship changed, asked about stories up to roughly the length of a long novel.
Fiction.live reports that today's models struggle with real long-context writing tasks.
Last updated 2026-04-04Covers 6 of the models we trackFiction.live offers AI tools that help writers make summaries, timelines and character bibles.Scores not shown: no license for reuse stated
See their resultsMore about this test: Fiction.LiveBench
Reliability not measured by this source.
Technical details for Fiction.LiveBench
Accuracy by context length: Questions about long stories that need real understanding, such as who knows a secret or how a relationship changed, asked about stories up to roughly the length of a long novel.
License: Score reuse not stated. Checked 2026-10-28.
What this doesn't tell you
- A model's advertised input length is not the length it can use well.
- None of these tests used your documents.
- Scores from different sources are not comparable, even when they share a unit.
- None of these tests were run on your own documents, code or data.
- No source for this task allows us to show its figures yet. Follow the links for their results.
Data version 2026-10-28+91ac3fc9d071 · Terms of use