Nemotron 3 Nano Omni 30B-A3B on EgoSchema
65.8%Rank #8 of 9 on EgoSchema
- Coverage
- Full (500/500)
- Cost per task
- —
- Median latency
- 12.8s
- Tokens per task
- 25,936 in · 2,959 out
Run setting: 1 fps.
Tokens and latency
Tokens per call
| Token type | Mean | Median | 90th pct. |
|---|---|---|---|
| Input | 25,936 | 25,898 | 26,075 |
| of which cached | 0 | 0 | 0 |
| Output | 2,959 | 1,281 | 7,701 |
Over 500 calls that returned an answer, including questions asked again after a failed attempt.
Latency
- Median
- 12.8s
- Mean
- 21.9s
- 90th percentile
- 44.5s
- 99th percentile
- 107.3s
We only consider the first question against a video for self hosted models, ignoring prefix caching latency.