Video-MME v2

Video understanding benchmark with 800 videos and 3,200 multiple-choice QA items spanning retrieval, temporal understanding, and complex reasoning.

4rows
scoreprimary metric
2026-05-06sampled

Metadata

ID
video_mme_v2
Category
Multimodal
Release
2026-04-06

Metrics

Video-MME v2 score

Latest Results

Rows are ranked by the Hugging Face leaderboard API rank. Model display names are preserved from source modelId values.

Rank Subject Video-MME v2 score Model Match Provenance Sampled
1 moonshotai/Kimi-K2.5 61.10 KIMI MoonshotAI: Kimi K2.5
moonshotai-kimi-k2.5
 		Imported 2026-05-06
2 Qwen/Qwen3.5-397B-A17B 55.90 Qwen3.5 397B A17B
qwen-qwen3.5-397b-a17b
 		Imported 2026-05-06
3 moonshotai/Kimi-K2.5 54.40 KIMI MoonshotAI: Kimi K2.5
moonshotai-kimi-k2.5
 		Imported 2026-05-06
4 Qwen/Qwen3.5-397B-A17B 48.90 Qwen3.5 397B A17B
qwen-qwen3.5-397b-a17b
 		Imported 2026-05-06