GPT-6 Astra on NExT-QA
90.7%Rank #1 of 9 on NExT-QA
- Coverage
- Full (4996/4996)
- Cost per task
- $0.114
- Median latency
- 5.6s
- Tokens per task
- 11,321 in · 15 out
Run setting: 1 fps.
Tokens and latency
Tokens per call
| Token type | Mean | Median | 90th pct. |
|---|---|---|---|
| Input | 11,321 | 9,608 | 22,557 |
| of which cached | 7,029 | 5,120 | 19,712 |
| Output | 15 | 4 | 38 |
| of which reasoning | 9 | 0 | 28 |
Over 5,004 calls that returned an answer, including questions asked again after a failed attempt.
Latency
- Median
- 5.6s
- Mean
- 6.0s
- 90th percentile
- 8.9s
- 99th percentile
- 14.6s
Per call, end to end including file upload and processing.