Meta's Code Llama AI Falls Short of ChatGPT in Coding Tests
-
Tested Meta's new open-source Code Llama AI on 3 coding challenges that ChatGPT handled easily. Code Llama failed 2 out of 3 tests.
-
Code Llama failed to write a functional WordPress plugin and gave code with errors. ChatGPT wrote working plugin code.
-
Code Llama fixed a bug in a string processing function correctly like ChatGPT.
-
Code Llama failed to diagnose root cause of a complex bug that stumped me. ChatGPT solved it.
-
Code Llama examines surface issues, doesn't analyze code structure and flow to find solutions like ChatGPT. More work needed.