Great video, dude. Thanks. Nearby i am trying to install llama localy and was disapointed... I dont know it took so long time to give me an answer... it is something near 8 - 13 seconds... I got a lot of hopes for this because needed it for my job)))
I am totally confused about the naming… so, Gemini 1.5 was released some weeks back to a few people. Is this now something new or did you only get access now?
@@engineerprompt understood. is the 1.5 pro in gemini advanced going to have the ability to upload folder like you presented in the video, thanks again, I assume today what we have in gemini advanced is not the 1.5 pro as I am not able to upload video or folders
Totally get that hassle. Sometimes, a quick refresh does the trick, but other times, it just hangs tight for a bit before chilling out on its own. Probably just the server getting slammed all at once.
I tested it too as i got the 1 million token as well. I forgot to take the screenshot of its replies but if it did, everyone would have been laughing. It sucks ass.
I tested Gemini Pro 1.5, and it's very bad. I'm not sure why people are so hyped about it. Probably just blinded by Google's research paper, which seems very off from reality, again 😥 In my tests, Gemini Pro 1.5 could not retrieve information from approximately 100k (yes, just 1/10th of 1M) of context tokens, so I think Google just messed up something with their evaluation of context retrieval with 1M tokens. It would definitely blow my mind if it worked as they promised, but in reality it just does not.
I think the context window is what's blowing peoples minds. It's almost an order of magnitude bigger than all other models and it's information retrieval is state of the art. But I agree that other than that the core model performs significantly worse than both Claude and ChatGPT right now. However I really love that it's set a new standard for context windows that GPT-5 kind of have to beat. All these incremental improvements are feeding into each other and is going to create something truly amazing very soon.
@@noone-ld7pt very bad as in ? The context window is cool but the main cool part is the multi modality. I have been using for extensively (for one day - just got access yesterday) and I can say that this is definitely on part gpt 4 turbo/claude 3, I'm going to say that it is even better than gpt 4 turbo in case of following instruction. Haven't tested the creativity aspect so can't say how does it compare to other models.
@@noone-ld7ptIn my tests, Gemini Pro 1.5 could not retrieve information from approximately 100k (yes, just 1/10th of 1M) of context tokens, so I think Google just messed up something with their evaluation of context retrieval with 1M tokens. It would definitely blow my mind if it worked as expected, but in reality it just does not.
If you are interested in leanring more about Advanced RAG Course, signup here: tally.so/r/3y9bb0
Great video, dude. Thanks. Nearby i am trying to install llama localy and was disapointed... I dont know it took so long time to give me an answer... it is something near 8 - 13 seconds... I got a lot of hopes for this because needed it for my job)))
Thanks 👍 A comparison of different AI coding assistents would be great ✌️
good suggestion. Will do.
Great video! Wish I could afford you on my project.
Email me, let's chat
I am totally confused about the naming… so, Gemini 1.5 was released some weeks back to a few people. Is this now something new or did you only get access now?
he only got access now
Yup, they are now giving more and more people access. Just got mine
Is gemini advanced the same than Google AI Studio, sorry I don't know how to access Google AI Studio
Ai studio doenst have the advance version. That is only for paid customers
@@engineerprompt understood. is the 1.5 pro in gemini advanced going to have the ability to upload folder like you presented in the video, thanks again, I assume today what we have in gemini advanced is not the 1.5 pro as I am not able to upload video or folders
Are you from India?
No
@@engineerprompt Where are you from? You sound a lot like my Indian Computer Science teacher.
I tried it. It says it's not Gemini but bard and he doesn't know about Gemini. It can't analise videos
I got access to the model but I hit an internal error in any of the models I use in AI Studio. Anyone else with this problem?
Totally get that hassle. Sometimes, a quick refresh does the trick, but other times, it just hangs tight for a bit before chilling out on its own. Probably just the server getting slammed all at once.
@@agnosticatheist4093 been days like this though. Couldn't use it even once.
Verify your default sign in google account is the same as the account with granted access to 1.5 pro api. Should be fine once this is verified.
@@Set_Your_Handle123 yeah. That was the problem. Genius. Thanks. 😎
I tested it too as i got the 1 million token as well. I forgot to take the screenshot of its replies but if it did, everyone would have been laughing.
It sucks ass.
how to get acces to this model?
Get on the waiting list. They are sending out invites now
Do you have to pay for this?
not right now. Its free for the time being.
It's said to support Polish language, but it's still not available in Poland. Are we some sort of third world country, or what?
yes. Its because Poland is a third world country.
European laws restrict many services. Talk to your beloved politicians.
It's available in india.
Well it disappeared from the map more than a few times in history, so...
@@xmarkclx 💀😂
I tested Gemini Pro 1.5, and it's very bad. I'm not sure why people are so hyped about it. Probably just blinded by Google's research paper, which seems very off from reality, again 😥
In my tests, Gemini Pro 1.5 could not retrieve information from approximately 100k (yes, just 1/10th of 1M) of context tokens, so I think Google just messed up something with their evaluation of context retrieval with 1M tokens. It would definitely blow my mind if it worked as they promised, but in reality it just does not.
I think the context window is what's blowing peoples minds. It's almost an order of magnitude bigger than all other models and it's information retrieval is state of the art. But I agree that other than that the core model performs significantly worse than both Claude and ChatGPT right now.
However I really love that it's set a new standard for context windows that GPT-5 kind of have to beat. All these incremental improvements are feeding into each other and is going to create something truly amazing very soon.
Yep just more trash from an emperor who has no clothes
@@noone-ld7pt very bad as in ? The context window is cool but the main cool part is the multi modality. I have been using for extensively (for one day - just got access yesterday) and I can say that this is definitely on part gpt 4 turbo/claude 3, I'm going to say that it is even better than gpt 4 turbo in case of following instruction. Haven't tested the creativity aspect so can't say how does it compare to other models.
@@noone-ld7ptIn my tests, Gemini Pro 1.5 could not retrieve information from approximately 100k (yes, just 1/10th of 1M) of context tokens, so I think Google just messed up something with their evaluation of context retrieval with 1M tokens. It would definitely blow my mind if it worked as expected, but in reality it just does not.
Please can you give a bit more concrete examples of what you found Gemini Pro 1.5 to be bad at?