In the openAI playground the output is of much higher quality than in a deployment, even with same parameters. What can I do

Question

When I test an Azure OpenAI deployment in the playground, everything goes smoothly. I test until i have the wanted result in the output. However, then I use the given code to do an HTTP request via PowerAutomate. All the settings and parameters exactly the same. I get a lot less better output. I don't think it has to do with the code/http/powerautomate, because I have the same issue when I deploy to web app via the playground.