Wrong Formatting in UI?
I'm using this model to chat in Open WebUI and it works just fine. However, the replies always show the reasoning part as well, together with some tags. Like when I simply say "Hi", the model replies with:
"<|channel|>analysis<|message|>We need to respond as ChatGPT: The user says "Hi". Probably we need to reply with a greeting and ask how can I help. Possibly ask for what they need. The user likely expects a friendly greeting. We'll respond: "Hello! How can I assist you today?"<|end|><|start|>assistant<|channel|>final<|message|>Hello! 👋 How can I help you today?"
Other models format the thinking process properly, just GPT-OSS shows these tags.
What do I have to change to separate the thinking process from the actual answer and omit the <|xxx|> tags?