Gemini 3.8 Flash on NExT-QA
88.0%Rank #3 of 9 on NExT-QA
- Coverage
- Full (4996/4996)
- Cost per task
- $0.0041
- Median latency
- 5.8s
- Tokens per task
- 3,904 in · 300 out
Run setting: 1 fps.
Tokens and latency
Tokens per call
| Token type | Mean | Median | 90th pct. |
|---|---|---|---|
| Input | 3,904 | 3,239 | 7,901 |
| of which cached | 1,307 | 0 | 7,855 |
| Output | 300 | 179 | 588 |
| of which reasoning | 298 | 177 | 587 |
Over 4,996 calls that returned an answer, including questions asked again after a failed attempt.
Latency
- Median
- 5.8s
- Mean
- 6.7s
- 90th percentile
- 9.5s
- 99th percentile
- 20.4s
Per call, end to end including file upload and processing.