Benchmark suite

Run the ImageBench V1 suite

Generate against a fixed 192-prompt suite from ImageBench, grade each output pass/fail with a VLM judge, and contribute to the live leaderboard.

Text Rendering5 tests

Spatial Reasoning19 tests

Human realism14 tests

Professional Studio9 tests

Graphical design8 tests

Truthfulness9 tests

Model

Pick a model and scope

Run the suite to grade outputs pass/fail. Results feed the live leaderboard.