Deepseek 3 greatly decreased speed..
Reda Izo
Hello there,
I wanted to share some feedback regarding Deepseek, which I've been actively using over the past few weeks, as my main LLM. I noticed that since the recent R1 update or other changes were implemented, the response speed has significantly decreased - it's now about 10 times slower than before. Previously, Deepseek's performance was comparable to GPT-4 or Claude 3.5 in terms of speed.
I'm curious to understand what might have caused this change in performance. Has anyone else experienced similar issues?
Thank you for taking the time to read this feedback. Looking forward to your response.
Best regards
R
Priyanshu
Hey Reda Izo Peter Karlen Reed Mayhew
Can you please try again? There were some issues with deepseek models that were identified and resolved. Let us know if you face any issues.
The responses might still seem a bit slow as deepseek api is down and other providers are also having issues. We are working with them to get this fixed
Thanks
Reda Izo
Priyanshu The situation hasn't changed unfortunately. I was under the impression that you guys were running open sourced AI models on your own infrastructure, but I now understand you're using DeepSeek's API services. This makes sense from both a business and financial standpoint i guess.
If DeepSeek is experiencing extremely high global traffic this week & having trouble keeping up with demand, there's not much we can do about it.
P
Peter Karlen
Yes, and it stops a few seconds into answering and I have to tell it to "continue" several times.
R
Reed Mayhew
Yes, and Deepseek 3 R1 is completely unresponsive at times. Please fix.