Back to today’s briefing

Story perspectives

Harvard's Llama 3.1 AI Matches GPT-4 in Diagnostics

3/31/2025

26 3

Summary

Story summary

Harvard Medical School researchers have unveiled that the open-source AI tool Llama 3.1 405B matches the diagnostic accuracy of GPT-4, offering hospitals a secure way to utilize patient data. This breakthrough could revolutionize research and model development, but it emphasizes the crucial need for human oversight in the AI diagnostic journey.

1 of 1

Story summary

Harvard Medical School researchers have unveiled that the open-source AI tool Llama 3.1 405B matches the diagnostic accuracy of GPT-4, offering hospitals a secure way to utilize patient data. This breakthrough could revolutionize research and model development, but it emphasizes the crucial need for human oversight in the AI diagnostic journey.