Hacker News Reader

Google's experimental thinking model is similarly casual. Not as casual as QwQ, but more casual than Gemini 1.5 Pro. Flash 2.0 will also go a bit more casual in its responses randomly, and when you tell it to think step by step.

I find OpenAI's models are the least fun to work with.

Yeah, both GPT-4 Turbo and GPT-4o are much less fun and creative than the original GPT-4.

Restating the question to start the response is good practice for Actual and Artificial Intelligence.