Basic Insight

We gave Claude and ChatGPT the same chart and asked a simple question.

The answers were... different.

In our latest benchmark, we tested how frontier AI vision models interpret a radial chart — the kind of complex visualisation we work with daily in market research. One model got it right. The other didn't — and no amount of prompt engineering, chain-of-thought reasoning, or even a custom Python CV pipeline could fix it.

The takeaway isn't that one model is "better." It's that reliable AI systems need multi-model architectures — the right model for the right task, with cross-validation built in.

Read the full write-up from our AI Research division 👇
https://marketing.ai.research.my/ai-eyes-disagree-benchmarking-llm-vision-radial-charts/

Insights' Privacy Day- a topic that everyone talking bout..