GPT-6 Astra on Q-Bench Video
68.3%Rank #4 of 9 on Q-Bench Video
- Coverage
- Full (868/868)
- Cost per task
- $0.0257
- Median latency
- 1.8s
- Tokens per task
- 2,434 in · 27 out
Run setting: 1 fps.
Tokens and latency
Tokens per call
| Token type | Mean | Median | 90th pct. |
|---|---|---|---|
| Input | 2,434 | 2,319 | 4,757 |
| of which cached | 13 | 0 | 0 |
| Output | 27 | 23 | 63 |
| of which reasoning | 19 | 13 | 53 |
Over 868 calls that returned an answer, including questions asked again after a failed attempt.
Latency
- Median
- 1.8s
- Mean
- 2.0s
- 90th percentile
- 3.0s
- 99th percentile
- 4.8s
Per call, end to end including file upload and processing.