Doesnt gpt-oss-120b perform better across the board at a fraction of the memory? Just specced a $4k mac studio that can easily run that at 128 gb memory.