Meta's Code Llama AI Falls Short of ChatGPT's Coding Abilities in Early Tests
-
Tested Meta's new open-source Code Llama AI against the same coding challenges that ChatGPT aced, but Code Llama struggled and often failed.
-
Code Llama failed to generate correct WordPress plugin code and header in first test.
-
Code Llama successfully rewrote string processing function in second test.
-
In third test involving a complex debugging challenge, Code Llama made superficial recommendations that didn't solve root issue like ChatGPT did.
-
Code Llama currently lags far behind ChatGPT's coding abilities, though it offers more privacy if self-hosted. Meta needs to improve it.