>>106230792
>Deepseek distilled their own model too, but their distills were shit.
it can never be overstated how shit
it's worse than the models they were based on in every way except for the most synthetic math maxxing benchmarks
it's brand damaging to do this and I never understood the point, a lot of retards download that "deepseek" from "ollama" and come out thinking it's retarded