Skip to content

LVBench

A long-video understanding benchmark for retrieving and reasoning over information distributed across extended video inputs.

6 models

Higher is better.

  1. 1Gemini 4 ArgonGoogle91.7
  2. 2Gemini 3.8 FlashGoogle87.1
  3. 3Gemini 3.7 FlashGoogle85.4
  4. 4Qwen3.8 MaxAlibaba81.8
  5. 5Qwen3.8-Omni-FlashAlibaba76.9
  6. 6Qwen3.8-Flash-NextAlibaba76.6
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Long-video understanding score
Tasks
Long-form video question answering
Authors
Qwen Team (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.