LVBench

multimodal
+
+
+
+
About

LVBench is a multimodal evaluation benchmark that tests AI models' ability to understand and reason about visual content in extended contexts. This benchmark evaluates models' capacity to process long visual sequences, maintain visual memory, and perform reasoning tasks that require sustained attention to visual information across extended multimodal inputs and complex visual narratives.

+
+
+
+
Evaluation Stats
Total Models5
Organizations2
Verified Results0
Self-Reported5
+
+
+
+
Benchmark Details
Max Score1
Language
en
+
+
+
+
Performance Overview
Score distribution and top performers

Score Distribution

5 models
Top Score
49.0%
Average Score
44.7%
High Performers (80%+)
0

Top Organizations

#1Alibaba Cloud / Qwen Team
3 models
47.2%
#2Amazon
2 models
41.0%
+
+
+
+
Leaderboard
5 models ranked by performance on LVBench
LicenseLinks
Feb 28, 2025
Apache 2.0
49.0%
Jan 26, 2025
tongyi-qianwen
47.3%
Jan 26, 2025
Apache 2.0
45.3%
Nov 20, 2024
Proprietary
41.6%
Nov 20, 2024
Proprietary
40.4%
+
+
+
+
Resources