I can also confirm this Bug fighting with it for two weeks already and this is making chatGPT 4 almost useless for my needs. At some point when generating code (in most of my cases) it gives a network error and even on retries, the network error persists. But if you log out and log in again, it shows as a complete response for the previously incomplete network error in chat. This is a huge bug (from the customer’s perspective) that needs to be fixed ASAP. If there were alternatives, I would already be elsewhere.
3 Likes
Same here. I can’t understand, why all connections are loaded in the US. Why is not a CDN used, for continents. If i use a VPN in the US, very fast response for a while. Then I changed the server, again very fast.
And code blocks have been corrupted since yesterday. Broken code blocks again and again (we had this error a long time ago, and I thought, this was solved)
And always truncated code, and we as humans should assemble functions manually via copy paste?
So strange. This results in repeating queries again and again. I would prefer, OpenAI prevents this and deliver complete code.
1 Like
I’m experiencing the same issue. I have GPTs to train on text PDF files and ask it to summarize and expand the ideas but it keeps returning Network error. Truly irritating!
2 Likes
Crazy stuff.
gpt4-1106-preview breaks the layout and witch between codeblock and normal output and move to naother languages like sql.
And in Playground, with GPT-4 is all ok, yesterday, i had sometimes also the same issue with broken layouts with gpt4-1106-preview. With the gpt4 model not.
And yes, very frustrating with the network errors.
1 Like
I’ve been constantly getting NETWORK ERRORS for the nth time and I can’t get any work done because of this. How to fix this? Is it a bug on your end or something?
2 Likes
I just upgraded this week and am quite annoyed at how often I am running into this network error issue. Is there a productive way to report this so the PMs get the right prioritization of this bug?
2 Likes
February 16th now and I’ve had this network error problem for at least a week, solid 90% of my interactions with GPT 4 ends up in network error. It mostly happens at the end of the answer… when its trying to close the response it just hangs and crashes to network error.
I’m getting this on every chat with an uploaded PDF. Even though the response is finished, I can’t continue the conversation because it doesn’t “complete” and there is no way to reply apart from clicking “regenerate” (which just fails again) or refreshing and editing a prior message (which just re-writes a new response, then fails again).
It also means I have no way to restart the conversation without losing all the context to this point, which is exhausting and is going to waste a ton of my chats quota.
Used to be great, fast, and reliable now it’s terribly slow, and always getting NETWORK ERROR!
Let’s be honest with ourselves here. @openAI was not architecturally prepared for millions of users.
To put it simply TOO MANY PEOPLE ARE USING IT
1 Like
Yeah, have the same issues.
I can’t understand, why OpenAI doesn’t use CDN. All traffic goes for my understanding in die US. From Europe, Asia… possible, it’s impossible?
There was a hint in the forum. i don’t remember to clear cache, change the language and login and something else. I can’t remember the steps and what is exactly to-do.
Im still having this issue. It started after o-1 was released. It is to the point where I can only use 4o because o-1 throws an error while it is still thinking.
I got this since yesterday… to a point where nearly every response breaks when it gets over max ~500 token (which can be less and mostly is less … so 4-5 words and it breaks with a network error)…
It’s getting into a ERR_NETWORK_CHANGED - while I am having the same IP address…
ChatGPT is unusable for 2 days now.
I resolved a similar issue before by logging out and using a VPN with a different IP from another country then logging back in again.
After that turning off the VPN didn’t change the positive outcome.
2 Likes
Tried that, configured a wireguard server with american IP address, I was connected to that, logged out of ChatGPT and in again, tried a couple of times (at least 10 times) and got the error every single time.
Even got a second error stating a model that should be used can’t be found… when I was trying to continue a chat or a starting a new one.
I checked my network with mtr and found out that I have pretty high packet loss when connected to vpn though (something with the tata network).
When I am not over vpn I have a very very very!!! stable internet connection with 0% packetloss and a ping to 8.8.8.8 of ~11-15ms…
I must add that after reloading the chat (after the time you would estimate the response would normally take) the response is shown there… Theoretically I could work like this… but still annoying as hell because I prefer to read the answer while it is produced to get a higher production rate.
Same computer different browser does not work (also network connection error).
But same ip adress other computer (not my account) and it works (at least the free version)…
So it might be a problem with my account or with the chatgpt/model version currently deployed to my account.
If that is the case - please remove me from the beta tester list.
2 Likes