So meta ai announced the code llama model which performs better than gpt3.5 and close to gp4 but the most interesting part is that it has 100k context which is a really huge thing specially in coding and having the context of a big code base.
I was wondering if you guys have any plan to add this as option or implement it into your platform or you guys focus on different stuff.
Very interested in adding Code Llama. We’re playing with it ourself to get a sense of its abilities.
If you end up playing with it, would love to hear your opinion on it’s abilities vs 3.5 and 4.
Here is a highly finetuned version of codellama that beats GPT4 at humaneval. Seems to be a better comparison than the stock one.