FrontierThe story, in brief
Even the best AI models lose about half their performance when charts get complicated, new benchmark finds
50% performance cliff. That's what happens when you ask leading AI models to read complex charts—new RealChart2Code benchmark exposes a critical weakness.
The KeyNews take
Why it matters
New benchmark reveals a significant capability gap in leading AI models when handling real-world complex visualizations, challenging assumptions about multimodal performance and exposing a potential liability for enterprise deployment.
The key facts
4 to knowRealChart2Code benchmark tested 14 leading AI models
Top proprietary models lose ~50% performance on complex charts vs. simpler tests
Test based on real-world datasets and visualizations
Performance degradation suggests brittleness in vision-language reasoning on structured visual data
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: The RealChart2Code benchmark puts 14 leading AI models to the test on complex visualizations built from real-world datasets. Even the top proprietary models lose nearly half their performance compared to simpler tests.