Not sure what to say. I mean, using the same model as you, this is what I got:
That said, not sure what you’re using on or if you watched setting. For example, see the toggle in screenshot below where it says to shape answers based on the page’s contents?
When I toggled that on, it then answered inaccurately as you can see below:
'50.36% of 2024 passed ’ is still not correct even toggle off the ‘based on the page’s contents’ and I started Leo AI at New Tab, not based on any contents.
Maybe Leo AI is created for summarizing the web page content.
Thanks anyway.~
1 Like
Well, there’s a lot of progress being made on things. I haven’t tested the other models on Leo and not sure if any of them might be accurate. Overall I’ve been using ChatGPT.
Speaking of, answer from ChatGPT on the same question:
Lmao, Meta is dumb too.
Claude 3 Haiku seems a little better, though I’ll have to pay attention to whether it or ChatGPT was more accurate.
Anyway, thanks for sharing. I moved the topic over to Brave Leo Feedback so the right people at Brave can see it.
You’re better off not using LLMs as calculators. Or expect errors even when they do give you an answer.