Skip to content

MMLongBench-Doc

A long-document multimodal benchmark for grounded reasoning over extended document contexts.

5 models

Higher is better.

  1. 1Nemotron 3.5 Super VL 120B A12BNVIDIA62.8
  2. 2Qwen3.5 397BAlibaba61.5
  3. 3Gemini 3 ProGoogle60.5
  4. 4Kimi K2.5Moonshot AI58.5
  5. 5Nemotron 3 Nano Omni 30B A3BNVIDIA57.5
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Document-grounded reasoning
Tasks
Long document understanding
Authors
Qwen (2026)

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