On an attempt to see how gemini behaves, I started experimenting with astrology and compared with Chat GPT and Prashna. Technically, I give the charts, cusps and ask for the answers. I don’t worry about the accuracy. But, on comparing with GPT5.2, following is happening:
- Gemini tries to super supportive prediction, while GPT 5.2 straight goes for honesty. This is even after instructions properly mentioned to prefer honesty. GPT 5.2 and Gemini always answers with differences.
- When I copy the answer of the Gemini 3 and give to GPT 5.2 and mention that there is different and ask it to update the answers, GPT 5.2 stays almost 80-90% same and maintains it’s answers clearly. I attempted it 92 times and GPT 5.2 changes its answers 3 and defended the answers with atmost proper justifications.
- Meanwhile… Gemini 3, when I paste answers from GPT 5.2 and tell gemini that there is difference in prediction, Gemini changed everytime. It says your answer is right and changes the answer.
What I mean to say is, Gemini 2.5 was not behaving like this. Meanwhile, Gemini 3 is easy to gaslight. It doesn’t focus on fact. It just focusses on giving the answer which user would be happy to hear.
Accuracy with predictions: GPT 2.5 Pro > GPT 5.1 Thinking > GPT 5.2 > GPT 5 > Gemini 3
Surprisingly, I did the same set of experiments when 2.5 was present and that time, it was super accuracte. It never changes its stand and always whatever it predicted, actually happened. That’s kind of mind blowing.