Molmo 2 8B on Q-Bench Video
62.9%Rank #9 of 9 on Q-Bench Video
- Coverage
- Full (868/868)
- Cost per task
- —
- Median latency
- 0.2s
- Tokens per task
- 918 in · 2 out
Run setting: 1 fps.
Tokens and latency
Tokens per call
| Token type | Mean | Median | 90th pct. |
|---|---|---|---|
| Input | 918 | 854 | 1,824 |
| of which cached | 6 | 0 | 0 |
| Output | 2 | 2 | 2 |
Over 868 calls that returned an answer, including questions asked again after a failed attempt.
Latency
- Median
- 0.2s
- Mean
- 0.3s
- 90th percentile
- 0.5s
- 99th percentile
- 0.6s
We only consider the first question against a video for self hosted models, ignoring prefix caching latency.