Azure openai Chatgpt-4o not accurate

Question

I used azure's openai service and deployed the model with the model name chatgpt4o, but through the chat it feels not the same as the free chatgpt4o on the chatgpt official website, I looped the version of it on the playground and what was returned wasn't even chatgpt4o or even it didn't know about chatgpt4o, it was azure's Is chatgpt4o not the same as openai's chatgpt4o?

I have seen other people have asked a similar question, azure people answered that chatgpt return not recognized just knowledge base issues and the model itself has nothing to do, but after I have chatted many times to compare, openai's chatgpt4o in answering the question when the language will be better, more like human and azure's is very obvious feeling is not the same underlying model. I think azure's 4o uses 4o-turbo.

Looking forward to your reply

Translated with DeepL.com (free version)

Answer 1

Hi there gl wz

Thanks for using QandA platform

Azure OpenAI’s ChatGPT-4o is not exactly the same as OpenAI’s ChatGPT-4o on chat.openai.com. yes,, both use the GPT-4o model, the underlying infrastructure, configuration, and tuning can vary between Azure and OpenAI’s own API.

Azure typically uses a variant of GPT-4o, like gpt-4o-turbo, optimized for cost, latency, and stability. These differences can impact how the model handles context, creativity, and reasoning.

If you're comparing free ChatGPT-4o with Azure’s 4o, it's possible that OpenAI’s own platform has additional tuning or features (like memory, system prompts, or fine-tuned behavior) that Azure’s version doesn’t expose by default. try experimenting with Azure’s system message or temperature settings to get closer to the OpenAI experience, but some differences may still be there but we can tweak make changes to get desired rrsults,

If this helps kindly accept the answer thanks much.