Skip to content

AI2D test split

A diagram understanding benchmark focused on scientific and educational visual question answering.

7 models

Higher is better.

  1. 1Gemini 3 ProGoogle94.1
  2. 2Qwen3.5 397BAlibaba93.9
  3. 3Qwen3.6-35B-A3BAlibaba92.7
  4. 4GPT-5.2OpenAI92.2
  5. 5Kimi K2.5Moonshot AI90.8
  6. 6Nemotron 3 Nano Omni 30B A3BNVIDIA88.5
  7. 7Claude Opus 4.5Anthropic87.7
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Diagram-grounded QA
Tasks
Diagram 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.