Gemini 1.5 Pro has a massive context window

[@artin_kim]

Really great video. Love the easy digestible format. Keep up the good work!

[@Daq0192]

Great video like always

[@ekstrajohn]

This is why I never got into RAG or LangChain. It's possible we will get this context lengths in open source next year, so what's the point? The only catch is you need a loooot of VRAM for it.

[@420_gunna]

50 cents per API call with Gemeni 1.5 Pro vs fractions of fractions of a penny for a RAG'd-out lil guy (and they can perform equally well for certain tasks, if the IR setup is goated) But yeah, I don't know how long that price difference will last.

[@Dart_ilder]

So.. WHY does it have such a long context? I was 100% expecting smth more like Mamba-MoE. But they just say that it is "transformer-based"