A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis

A decade-old convolutional neural network (CNN) outperformed two of the most advanced large language models—GPT-6 Astra and Claude Fable 5.1—at diagnosing skin cancer, correctly identifying the diagnosis in 62% of 200 biopsied lesions compared to 42% for each LLM. The CNN, based on Google's 2015 Inception v3 architecture and fine-tuned on public dermatology images for about twenty minutes, achieved superior performance despite being intentionally not state-of-the-art, suggesting specialized deep learning models remain more effective than frontier LLMs for specific medical imaging tasks.