Video-MME: Results as Published

Video-MME tests whether a multimodal model can analyze video content across a wide range of clip lengths. The set holds 900 videos totaling 254 hours, with durations from 11 seconds to 1 hour, and 2,700 question-answer pairs. All data was newly collected and annotated by humans rather than taken from an existing video dataset. Model cards print Video-MME (w. sub), the setting in which the model also receives the subtitles matched to the sampled video frames; the score is accuracy, broken down by video duration, domain, subcategory and task type.

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.