Skip to content
Vision Index

Kimi K3 on Video-MME v2

About Video-MME v2How this is measured

55.2%Rank #5 of 9 on Video-MME v2
1,765 of 3,200 questions answered correctly.
Coverage
Full (3200/3200)
Cost per task
$0.275
Median latency
100.4s
Tokens per task
68,204 in · 4,721 out

Run setting: 1 fps.

Tokens and latency

Tokens per call

Token typeMeanMedian90th pct.
Input68,20468,27668,472
of which cached53,57367,58467,584
of which frames or video66,67166,75066,750
Output4,7212,13513,238
of which reasoning4,7052,11913,222

Over 3,281 calls that returned an answer, including questions asked again after a failed attempt.

Latency

Median
100.4s
Mean
210.6s
90th percentile
515.6s
99th percentile
1503.9s

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