Skip to content
Tech News
← Back to articles

Dermatologist finds older Inception v3 CNN beats GPT-6 Astra and Claude Fable 5.1 on skin cancer diagnosis

read original get DermLite DL4 Dermatoscope → more articles
GoKawiil Brief

A dermatologist tested GPT-6 Astra and Claude Fable 5.1 against a fine-tuned, decade-old Inception v3 convolutional neural network on 200 biopsy-confirmed skin lesions from the public MILK10k dataset. The CNN correctly identified the diagnosis in 62% of cases, versus 42% and 43% for the two frontier large language models, including correctly flagging a melanoma that both LLMs misclassified as benign and recommended for discharge.

Why It Matters

GoKawiil's interpretation of the reporting above, not reported fact.

The result suggests that general-purpose multimodal LLMs, despite their broad capabilities, may not yet match specialised, narrowly trained models for high-stakes visual diagnostic tasks like cancer detection. It implies that clinical deployment of AI for dermatology could benefit more from purpose-built image classifiers than from general frontier models, though this is based on one dermatologist's informal test of 200 cases rather than a peer-reviewed clinical trial.

Key Takeaways
Worth a Look

DermLite DL4 Dermatoscope — The article highlights how dermoscopic imaging was key to correctly identifying the melanoma that AI models missed, underscoring the value of quality dermoscopy tools. The DermLite DL4 is a widely used handheld dermatoscope that lets you capture the same kind of detailed skin images discussed in the piece. It's a great way to get closer to the clinical-grade imaging that made the difference in this diagnostic comparison.

See DermLite DL4 Dermatoscope on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.

Source: lynchlab.net — Dr Magnus Lynch Is A Consultant Dermatologist, 2026-10-25

Published there as: “A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis”

Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.