Posted by 737max 6 hours ago
This is crazy if it can work accurately.
Related funny https://www.youtube.com/watch?v=DlTNN0gvkLM
Pretty misleading.
It thinks a ton, ~24k output tokens a run vs ~3k for Sonnet. It missed a solution by thinking its way into a 10 min timeoout.
Something to look out for - if you're using pi, it doesn't have Argon in its catalog yet and reports cost at Kimi's rates, about a third of what OpenRouter actually bills.
Me: Gemini, navigate to nearest McDonalds.
Gemini: [Lists 10 McDonalds locations wnd asks which I want.]
Me: What were the priginal instructions just now?
Gemini: [Recites imperatives about tone and content from its memory system.]
Me: No. I told you to navigate to the NEAREST McDonalds. There's only one NEAREST McDonalds. Do that one.
Gemini: My mistake, navigating to the nearest McDonalds. [Opens Maps, navigating to a McDonalds across town.]
I also have had shockingly bad results with gemini on android. Ended up giving up on it after a few months. Silly silly google with their trademark UX polish and attention to usability.
Flash lite is very noticeably more idiotic than 3.8 Flash. It's why the quick AI overview is often much worse than the AI mode one.
The only thing I can figure is that Google's consumer-facing free-to-use models are set to "reason as fast as possible with only the most basic plausibility checks." I have to correct them like this every single time.
Similarly, you can run local LLMs on your Google Pixel smartphone, but if you want them to be hardware accelerated using the chip that's included for that very purpose on your phone... it has to be Google's models via Google's apps. So you can't say "alright, a fast, local model which sounds great, but I want one that is trained to always run a search before answering, in a harness with good search tools for answering the kinds of practical questions I often ask", you're actually not free to put that together in a practicable way.
So there is indeed some competition here. But it's not really fair competition.
1. I ask it to list out all my old Keep notes that are just throwaway task lists, reminders, etc. to propose deleting them.
2. It comes up with like 4 of them (I know for a fact there are dozens, if not hundreds). I ask it for an exhaustive list.
3. I see its thinking trace briefly show "Refusing to produce an exhaustive list..." and then it says "Yes, that is all of them."
Not sure whether it's lack of intelligence, bad tools, or both, but that was extremely frustrating to see.
If it had exact location access people would be crying that Gemini steals their location data.
I think these companies have good models but expose them poorly like others have mentioned
I tell it to "Play Let it go from frozen". It plays it on youtube music. 5 seconds later I tell it again 'Play let it go from frozen" it then just dies saying "I can't play videos in car".
i benchmarked 20+ models in real worlds and gemini families are the best (highest customer satisfaction)