I'm curious about the differences between DeepSeek R1 0528 and the original R1 model. Are there any significant improvements or issues that you've noticed with the new version? I'd love to hear about your experiences with it!
5 Answers
I’ve tested math functions, and only DeepSeek R1 0528 got tricky problems like 9.9-9.11 right. It’s impressively accurate!
It feels like R1 0528 has the smarts of the old R1 but the composure of V3. The first R1 behaved unpredictably, but this new version is much more controllable and reliable!
From what I’ve seen, it returns shorter responses—about 60% less lengthy—and uses 50% more elaboration. Plus, I’ve noticed it goes crazy with emoticons—three times more than before!
Hmm, interesting point!
One major upgrade is that it can now think in multiple languages, not just English. It even passed my native language poetry test, which is a big step forward!
It was able to handle Russian in the previous versions; good to see it's retaining that!.
Yo mama!
ChatGPT free gets it right all the time, though!