79.7%
▲ 19.8 pts since 2023
best Video-MME score, video-SALMONN 2+ (2025)
20232025
What the chart shows
Video-MME tests multiple-choice questions about short, medium and long videos. The best score by model release year rose from 59.9% in 2023 (GPT-4V), to 75% in 2024 (Gemini 1.5 Pro (Feb 2024)) and 79.7% in 2025 (video-SALMONN 2+).
Long videos are the hardest part: models lose track of details spread across up to an hour of footage.
Key numbers
| Measure | Value |
|---|---|
| 2025: video-SALMONN 2+ | 79.7% |
| 2024: Gemini 1.5 Pro (Feb 2024) | 75% |
| 2023: GPT-4V | 59.9% |
How to read it
- Results collected by Epoch AI; the overall score without subtitles, so the model must follow what it sees and hears.
- Each bar is the best result among models released that year, from Epoch AI’s benchmarking hub. Scores depend on the model version and settings; years are model release years.
Source: Epoch AI, 2026: Benchmarking Hub, read 3 November 2026. Chart and text by AI Adoption.
Numbers checked against the published data on 3 November 2026.
Download the data (CSV) · Cite this chart
Numbers checked against the published data on 3 November 2026.
Download the data (CSV) · Cite this chart
Discussion