The more recent Chinese models are no longer heavily limited by context size. It can easily fit in RAM on a prosumer laptop. (You can also use swap space to extemd that, since context is only written to once per inference, thus a relatively mild wear-and-tear concern.)