AI2D test split

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

ModelsModels with a result.8
Top resultBest result on this benchmark.—
Top-3 spreadPoints from first to third.—
YearYear of release.—

Result and price

6 of 8 results

Results

8 results
#ModelResultThe score from the source.
—Gemini 3 Pro94.1%
—Qwen3.5 397B93.9%
—GPT-5.292.2%
—Kimi K2.590.8%