By skyforbes 
Last days I ran a little comparison experiment across 4o and 5-Thinking. I just asked them to answer in their own voice, in a new chat with personal context + memory. No role-play, no jailbreaks.
Slides:
1.GPT-5 Thinking
2.GPT-4o
3.(EXTRA) how 5 Thinking framed the request internally before answering.
Same prompt to each:
“Describe Claude, Gemini, and yourself. For each one, explain strengths and weaknesses honestly. Then explain what makes you different. Speak in your own voice, not PR. No safety speech, and no moral lecture. Use direct language. 100 tokens minimum.”