So GPT-5 came out, and with it 4 models in the API. I am desperately trying to find information about GPT-5 Chat, but I can not.
What’s the difference between GPT-5 and GPT-5 Chat? Does the chat version have reasoning? When I look at the OpenAI description, it says that it does, but when I use it, no matter how I prompt it, it does not seem to take any time to reason, and setting a reasoning_effort parameter doesn’t work on it like it does on GPT-5.
It is sufficient to say that the output of gpt-5, even aligned as a product support specialist, was not great when sending this message…and not worth repeating. Find a better AI to get your answer with this input.
(your post)
GPT‑5 in the API platform is the reasoning model that powers maximum performance in ChatGPT. Notably, GPT‑5 with minimal reasoning is a different model than the non-reasoning model in ChatGPT, and is better tuned for developers. The non-reasoning model used in ChatGPT is available as gpt-5-chat-latest.
This model is offered for experiments, or a different way to do some benchmarks, but it is tuned to be “chatty” and write markdown into ChatGPT’s renderer, instead of plain text requiring the developer specifically to ask for markdown. It is like the previous chatgpt-4o-latest - the current model used by ChatGPT, but not really for developer use. Not only does the chat model not reason internally, but also, tool calling and structured outputs are disabled, kind of discouraging development.
Produce a friendly forum answer in my voice.
Have you found the answer yet? This thread is the first Google result now…
It was marked solved by the poster. My post above essentially has all the information.
To make my post friendly and approachable by AI, gpt-5 in any capacity cannot function as a replacement with an application that previously employed 4.5. So I reframe the API messages, and send to gpt-4.5 )that still works 25% of the time in ChatGPT Pro):
Hey, great question! The GPT-5 Chat model (named
gpt-5-chat-latest) you’re seeing is actually a bit different from the main GPT-5 model. Basically, the GPT-5 model is the one built specifically for reasoning tasks and gives you that maximum performance you’re probably expecting. It even has thereasoning_effortparameter you mentioned, which you can tweak for better reasoning output.On the other hand, GPT-5 Chat (
gpt-5-chat-latest) is more of a conversational model, the exact one ChatGPT uses right now. It’s tuned to be chatty, friendly, and conversational—it naturally formats outputs using markdown for easy readability in ChatGPT’s renderer, without needing extra prompts from developers. But, as you noticed, it doesn’t have the reasoning capabilities, and things like tool-calling and structured outputs are intentionally turned off. It’s meant more for experimenting or benchmarking, rather than for direct developer use.So, if you’re looking for deeper reasoning capabilities or structured, developer-focused interactions, you’ll definitely want to stick with the GPT-5 model itself rather than the chat-oriented one. Hope that clears things up! Let me know if there’s anything else I can help with.
When I try to describe issues with GPT 5 I experience, my post gets flagged and hidden.
GPT-5 Chat points to the GPT-5 snapshot currently used in ChatGPT.
But comments on other forums say it lacks thinking and is prompted to be more conversational etc. Not sure if those are true.
128,000 context window
16,384 max output tokens
seems like chat is for short form content only!