← Back to context Comment by szatkus 2 hours ago Most likely those models would be extracted from the hardware in no time. 1 comment szatkus Reply amelius 1 hour ago I suspect no, because most of the tokens would be used for its internal reasoning loop, and a cheaper model could be used to obfuscate the output.
amelius 1 hour ago I suspect no, because most of the tokens would be used for its internal reasoning loop, and a cheaper model could be used to obfuscate the output.
I suspect no, because most of the tokens would be used for its internal reasoning loop, and a cheaper model could be used to obfuscate the output.