Right here, for context, is what I mentioned:
I wish to present you ways straightforward it’s to check deepseek-r1. It’s a single command. I do know nothing about DeepSeek, find out how to set it up. I don’t wish to. However I wish to get my palms on it in order that I can perceive it higher. RamaLama may help!
Simply sort:
ramalama run ollama://deepseek-r1:7bWhen the mannequin is completed downloading, sort the identical factor you typed with granite or merlin and you’ll evaluate how they carry out by their outcomes. It’s attention-grabbing how DeepSeek tells itself what to incorporate within the story earlier than it writes the story. It’s additionally attention-grabbing the way it confidently says issues which might be improper 🙂
What DeepSeek thinks
I included in my electronic mail the outcomes of a question I entered into DeepSeek. I requested it to write down a narrative a few sure open source-forward software program firm.
DeepSeek returned an attention-grabbing narrative, not all of which was correct, however what was actually cool was the way in which that DeepSeek “thought” about its personal “pondering” — in an eerily human and clear approach. Earlier than producing the story, DeepSeek—which, like OpenAI o1, is a reasoning mannequin—spent a number of moments muddling by means of how it might put the story collectively. And it confirmed its pondering, 760-plus phrases’ value. For instance, it reasoned that the story ought to have a starting, a center, and an finish. It needs to be technical, however not too technical. It ought to discuss merchandise and the way they’re being utilized by companies. It ought to have a optimistic conclusion, and so forth.
