MEDit-Bench: leaderboard
MEDit-Bench tests message-driven video editing: given a long video and an editing message, a model must choose the sequence of cuts that tells that story, and its edit is compared with the cuts professional editors made for the same message. This board is the mean temporal intersection over union between the footage the model selects and the footage the editors selected.
Metric: mIoU (%; temporal IoU between the union of the predicted cuts and the union of the reference cuts, averaged over the 540 video, message and reference-edit triplets (60 long-form videos, three editing messages each, three professional edits per message); zero-shot: the model maps the video and the message to a sequence of cuts). Source: arxiv.org. Saturation forecast: Around 2032. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 25.3 | #167 |
| 2 | Gemini 3 Pro | 23.5 | #137 |
| 3 | GPT-5 Mini | 18.6 | #348 |
| 4 | Gemma 4 E4B (Non-reasoning) | 16.9 | #2214 (Gemma 4 E4B) |
| 5 | Gemini 2.5 Pro | 15.2 | #266 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=medit-bench · How It Works · Data refreshed daily, snapshot 2026-11-13.