Skip to content
67.5%Rank #6 of 9 on Q-Bench Video
586 of 868 questions answered correctly.
Coverage
Full (868/868)
Cost per task
$0.0156
Median latency
11.3s
Tokens per task
2,759 in · 485 out

Run setting: 1 fps.

Tokens and latency

Tokens per call

Token typeMeanMedian90th pct.
Input2,7592,6065,422
of which cached2700
of which frames or video2,5622,4035,180
Output4852161,070
of which reasoning4682001,054

Over 868 calls that returned an answer, including questions asked again after a failed attempt.

Latency

Median
11.3s
Mean
21.6s
90th percentile
46.1s
99th percentile
174.6s

Per call, end to end including file upload and processing.