1 post tagged deepseek.
Ran both models through the same set of coding and reasoning tasks. Results were closer than expected.