Kimi k3 repsonses degraded?

Is it just me or kimi k3 gives degraded and very fast resposnes after sometime. It was fine initially but now at midnight it’s giving super fast and very degraded resposnes that are far from the previous well articulated long responses. Is it a midnight thing or is the model degraded

IDK but right now the model is returning 404 for me.

I’m having the exact same problem, Kimi k3 keeps returning 404.I’m having the exact same problem, Kimi k3 keeps returning 404.

Pretty sure its down right now or something.

Glad I’m not the only having this problem. A few hours ago, I had two successful responses after the 24-hour 429 soft-lock but now this…

I guess nvidia is fixing its slow tps issue.

I suppose so, but DeepSeek is also getting the 404 error. I think Nvidia has a serious bug right now.I suppose so, but DeepSeek is also getting the 404 error. I think Nvidia has a serious bug right now.

I am also getting constant 404 rn I just got a good repsosne sowmtime ago

It appears it works now.

Edit: It’s intermittent.

Same with Minimax M3

Yeah for response in between now getting a combination of 404 429 and 410

Now i am getting “not found” error or just an endless response wait whenever i try to use NIM.

Anyone else?.

@b_casco Yeah, it happens to me too. It’s like the model was stuck on ‘Not Found’ for hours, then came back, and now it just keeps waiting for a response before eventually giving the ECCON error.

You could say that NIM is shut down; there are no working models available right now.

It’s pretty unstable. I got a response some minutes ago, but now it’s back to 404 in every retry.

Now apparently is fixed and they added a new Deepseek V4 Pro model!

@b_casco yeah, I’ve been testing it. What I did notice, though, is that DeepSeek Flash 0731 is really slow to respond. It barely gives any responses, so it seems unstable.

Is it fixed? I am still getting error 410 for kimi

I (again) got soft-locked by 429 despite making requests to K3 between 4-5 minutes each one. I don’t know if the model has a certain RPD or it’s just me who’s unlucky. I would really appreciate it if the API documentation showed the number of requests allowed to the models, this would save a lot of headaches. See you in 24 hours.

Managed to get 3 kimi responses after whole morning of getting error 410 on it and now getting error 410 and 429 again. I am so confused