rt, using my own test question: 似乎是又一个可以难倒ai的难题（多模态和综合工具调用）
Although 3.7 Flash went on at length, its judgment was correct (with hints to assist its judgment).
However, without any hints, it still made the same mistake as Pro.
It seems the Gemini model’s general knowledge still isn’t quite up to par! (Even though it’s supposedly the world’s strongest multimodal model with the most extensive knowledge.) 
Apparently, there currently isn’t a model that can pass on the first try without hints.
Forget it, I’ll give you an upvote—it hasn’t been easy for you.
You’re in every thread farming posts—what are the requirements for leveling up?
How could that be? Am I that kind of person? I always upgrade properly and take it seriously! You can’t say that about me, no way~
You can try out Anti-Gravity and V—the official website has the most comprehensive information. 
Are the evaluation questions you’re using more focused on commonsense reasoning or fine-grained image recognition? I feel like nowadays, small-parameter multimodal models often sacrifice knowledge storage for inference speed, so it’s pretty common that their commonsense capabilities don’t differ much from the larger-parameter versions.
Give AI Studio a try. Although Gemini’s world knowledge and multimodal capabilities are resting on their laurels, they’re still very strong.
Should it be?
They’re all there, right?
Does that count as comprehensive testing?
Hajimi is updating too slowly.
Why not try antigravity? The web version isn’t even as good as Ask Gemini on Chrome.
This is no longer just a matter of visual ability, is it? At this point, it feels like AI is about to break into the realm of “hardware.”
Isn’t the latest pro version still 3.1? Where did this 3.5 pro come from?
az, I typed it wrong by accident 
hyw
The visual capabilities are actually pretty good, but everything else really isn’t that useful 
I just want to be the Water King!
That’s incredible.
What do you mean? Do you also want to be put on the pillar of shame?
Don’t want to.
