deep down, i suspect they're all just drafting on implementations to llamacpp.