I’m reading through the Github Copilot update announcement, where they mentioned that they are using a new upgraded Codex model that “delivers better results for code synthesis”.
Is this internal only to them, or will we be seeing a new version of this model soon? I’m asking because I noticed that it does lack when it comes to code synthesis, so an upgrade that we can access would be very welcome.
It is impossible to know what model the announcement refers to, given that Microsoft has early access to all OpenAI LLMs. Maybe it refers to the code-davinci-002 which is newer and has double the context length compared to previous davinci-codex model, or maybe it’s a fine-tuned version of it, or maybe it’s a model that isn’t even announced yet.
