Video-ChatGPT: benchmark results
Provider: Other.
Aggregate Rating 1402 ± 34, rank #6022 of 10916 rated models, from 105 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| TOC-Bench - Event Count | 30.7 | Event Count accuracy (%) on TOC-Bench (2,323 human-verified | 95 |
| OpenVLM Video - MLVU - Action Count | 34.5 | Accuracy (%) | 65.3 |
| MLVU-Test - Ego Reasoning | 32.1 | Multiple-choice accuracy (%) | 48.5 |
| MLVU-Dev - Generation | 3.9 | Generation score, GPT-4 judged (0-10) | 48.3 |
| TOC-Bench - Reappearance or Disappearance | 47.8 | Reappearance or Disappearance accuracy (%) on TOC-Bench (2,3 | 47.5 |
| MLVU-Test - Needle QA | 28.3 | Multiple-choice accuracy (%) | 47.1 |
| MLVU-Test - Generation | 3.64 | Generation score, GPT-4 judged (0-10) | 46.9 |
| OpenVLM Video - MVBench - Egocentric Navigation | 34 | Accuracy (%) | 44.4 |
| VidPair-Halluc | 24.58 | Binary-QA weighted accuracy wAcc (%): question-pair and vide | 35.7 |
| MLVU-Test - Action Order | 17.1 | Multiple-choice accuracy (%) | 30.9 |
| MLVU-Test - Sports QA | 27.8 | Multiple-choice accuracy (%) | 30.9 |
| OpenVLM Video - MVBench - Action Antonym | 64 | Accuracy (%) | 27.8 |
Interactive version: theaggregate.ai/model?slug=video-chatgpt · How It Works · Data refreshed daily, snapshot 2026-11-08.