GPT-5 Pro and Coding in ChatGPT Never Completes

Trying to experiment with using GPT-5 Pro in ChatGPT for some coding tasks that include uploading some files.

I’m finding that consistently, if files are uploaded, it never completes. I have a few chats that are 12+ hours on the progress bar, fully filled but no response. CoT sidebar shows nothing.

Just me?

12 Likes

I have the same problem, Today I launched new chat with gpt-5 pro model but it never finishes coding tasks if files are attached.

Sorry to hear you’re having trouble but I am sort of glad it’s not just me… Let me know if you find any workarounds!

I had some similar problems with o3-pro when it was released, but after some time it was fixed. However, if I try to pass the same task now to o3-pro it completes it without problems.

For me I can reproduce this at will. I opened a support case, we’ll see.

1 Like

same issue, got any solution?

Sorry to jump into the conversation like this, but I’m planning to get ChatGPT Pro for coding and system building. I haven’t found good resources comparing GPT-5 Thinking and GPT-5 Pro. From your experience, is the Pro plan worth it?

“Worth it” is often tough not knowing someone else’s circumstances or what tradeoffs might be required to afford it so I can only speak for myself and I can put it simply - right now my Pro account is via my employer so I am not paying directly.

If that were to change, this subscription is one of the first things I would take on personally. I use it every single day!

Back on the original topic, I have been working on this with OpenAI support. Not resolved but I am to send them more information later today.

1 Like

Same issue. Look forward to your updates @hunter.hillegas.

Will update when I have more. They had me do a bunch of network tracing so I presume it’s being kicked up a level or two, may take some time.

2 Likes

Echoing the issue so it gets fixed - missing out on a ton of productive time waiting on the model to complete; and if it does complete the request, the download link it provides is empty

@Alton, I do have a workaround for the empty download link, which happens to me quite frequently in chat. Ask CGPT to give you a txt file that contains a single word. That is always successful, and I click to download it. Then ask CGPT to re-post the original download, and it is no longer empty.

2 Likes

Same issue… Thanks for starting this thread.

I don’t know if OAI works this way but if others are seeing the same thing, could be worth opening your own support incidents (referencing this thread) so they have a better sense that it’s impacting more people.

Same. I was hoping it was just doing a thorough job analyzing for hours.

Joining in, I have been having the same issue, I have had one or two complete (but sent in 10+) coding prompts to GPT-5 Pro. They are all still “Reasoning” 1+ days later. Additionally, the ones that did complete were extraordinarily off from the right solution, outputting code that does not even compile.

1 Like

I’m having the same issue with Pro :frowning: . Hoping to get a resolution soon

I suggest calling GPT-5 high in VS Code Copilot Agent Mode to work with your coding situation.

As for the Pro plan, I’m using a Teams plan that gets enough Pro usage that I can effectively use it with Thinking to get excellent outputs.

So, what I do, @hunter.hillegas, is figuring out what I want to do with Pro in ChatGPT. We’ll figure out a tight scaffold and architecture in a few prompts that’s extra crisp, ‘cause its Pro. Then I’ll take the refined work to GPT-5 in VS Code (or Cursor) and work with the actual changes there. It is far more forgiving, tuned to work in small steps, and you get in a rhythm for when to okay a move or something.

There are a couple other strategies for getting both versions hip to your Codebase.

I have the same issue. It also freezes on reasoning and cannot be paused, so technically I cannot continue the same conversation. So far the my solution is to submit the larger files and have them initially processed by another model, then switch to O5 pro.