LongBench: Results as Published

LongBench v2 tests whether language models can reason over very long inputs, with contexts from about 8,000 to 2 million words. It has 503 multiple-choice questions in six categories: single-document and multi-document question answering, long in-context learning, long dialogue history, code repository understanding and long structured data. Human experts scored about 54% under a 15-minute limit, which is the reference point for model scores.

Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.