I've been using DeepSeek for a while now for college assignments since its DeepThink feature really improved its performance compared to ChatGPT. Recently, though, I noticed some changes. It seems like it's not producing those in-depth, thinking responses anymore—instead, it just gives standard answers that aren't as helpful as they used to be. For instance, it can't generate assembly code like it did before. Has something changed with the DeepThink model recently? Is there any way to revert to the older version?
3 Answers
The latest update changed the wording from 'he-user' to 'we-us.' I think it’s part of the shift to a more user-friendly approach.
I felt the same way! Even if it seems ‘thoughtful,’ the language is less technical and more relatable now. It feels like it’s trying to be more empathetic, you know?
I’m not totally sure what v128 is, but DeepThink still seems to work for me. I had a chat with it yesterday, and it produced pretty solid responses. Check this out—looks like it’s still thinking!
I wish mine worked like that! Sometimes I get responses that seem sophisticated but then turn into simple answers. It's frustrating!
Do you know when this update happened? I'm curious about which version worked better before!