Has anyone else been having an issue where the models would answer its own question? I have been seeing this quite a lot with Gemini 4 models.
The "long-horizon" ones are optimized for this. They won't get a high score on the benchmarks if they stop to ask questions.
Oh yeah, and you're paying for it to say "wait. what about [[thing]]?" then answer itself.
The "long-horizon" ones are optimized for this. They won't get a high score on the benchmarks if they stop to ask questions.