Molmo2 Human Preference - Captioning: leaderboard
Captioning part of the Molmo2 human preference study (Allen Institute for AI and University of Washington, 2026): annotators compare two models' dense captions of the same video, over 51 videos. Scores are Bradley-Terry Elo ratings (bootstrap median) relative to the models in the study; higher is better.
Metric: Elo rating (bootstrap median, Bradley-Terry; pairwise human preferences on dense captions of 51 videos). Source: arxiv.org. Saturation forecast: Around January 2027. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 1148 | #262 |
| 2 | GPT-5 | 1136 | #163 |
| 3 | Gemini 3 Pro | 1126 | #135 |
| 4 | Gemini 2.5 Flash | 1109 | #452 |
| 5 | GPT-5 Mini | 1086 | #349 |
| 6 | Molmo2-8B | 1049 | #1488 |
| 7 | GLM-4.1V-9B (Thinking) | 1013 | #1061 (GLM-4.1V-9B) |
| 8 | Claude Sonnet 4.5 | 1009 | #244 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=molmo2-human-preference-captioning · How It Works · Data refreshed daily, snapshot 2026-11-12.