Molmo 2 8B on Video-MME v2
27.5%Rank #9 of 9 on Video-MME v2
- Coverage
- Full (3200/3200)
- Cost per task
- —
- Median latency
- 201.2s
- Tokens per task
- 29,699 in · 2 out
Run setting: 1 fps, 338-frame cap (673 videos thinned).
Tokens and latency
Tokens per call
| Token type | Mean | Median | 90th pct. |
|---|---|---|---|
| Input | 29,699 | 30,473 | 30,633 |
| of which cached | 22,137 | 30,320 | 30,336 |
| Output | 2 | 2 | 2 |
Over 3,200 calls that returned an answer, including questions asked again after a failed attempt.
Latency
- Median
- 201.2s
- Mean
- 189.3s
- 90th percentile
- 203.3s
- 99th percentile
- 205.3s
We only consider the first question against a video for self hosted models, ignoring prefix caching latency.