Qwen3.5-35B-A3B on UCF101-AD
41.3%Rank #6 of 9 on UCF101-AD
- Coverage
- Full (4224/4224)
- Cost per task
- —
- Median latency
- 4.3s
- Tokens per task
- 1,016 in · 1,176 out
Run setting: 1 fps.
Tokens and latency
Tokens per call
| Token type | Mean | Median | 90th pct. |
|---|---|---|---|
| Input | 1,016 | 872 | 1,265 |
| of which cached | 0 | 0 | 0 |
| Output | 1,176 | 638 | 2,567 |
Over 4,224 calls that returned an answer, including questions asked again after a failed attempt.
Latency
- Median
- 4.3s
- Mean
- 7.8s
- 90th percentile
- 16.7s
- 99th percentile
- 41.1s
We only consider the first question against a video for self hosted models, ignoring prefix caching latency.