Comment by m3kw9

Now do a programming task that requires more than 32k of context and see who’s “better”. If you don’t bench mark that you cannot get an overall pic. GitHub copilot for example could benefit big from the increased context