In my experience it’s not, the only difference I noticed between it and Opus was its taking much more time to respond.
If you think there is ever going to be an open source Deepseek "AGI" model I just don't think that is thinking things through.
It is the main error of the AI bubble. At some level of intelligence, the dual use nature of a model is too dangerous for a purely hands off approach.
It is like thinking at the advent of the automobile that you will be able to drive your car at any speed, without a license , any place you want.
It is inevitable and the huge sums of money being burned to build these future highly regulated public utilities probably aren't going to be happy with the returns they get from funding a highly regulated public utility.