Molmo 2 8B on NExT-QA
89.2%Rank #2 of 9 on NExT-QA
- Coverage
- Full (4996/4996)
- Cost per task
- —
- Median latency
- 1.0s
- Tokens per task
- 3,910 in · 2 out
Run setting: 1 fps.
Tokens and latency
Tokens per call
| Token type | Mean | Median | 90th pct. |
|---|---|---|---|
| Input | 3,910 | 3,268 | 7,895 |
| of which cached | 3,428 | 2,944 | 7,664 |
| Output | 2 | 2 | 2 |
Over 5,365 calls that returned an answer, including questions asked again after a failed attempt.
Latency
- Median
- 1.0s
- Mean
- 1.9s
- 90th percentile
- 5.3s
- 99th percentile
- 5.9s
We only consider the first question against a video for self hosted models, ignoring prefix caching latency.