Spaces:
Running
on
Zero
Not as introduced (just another AI)
author delete that space fr, that's not even close to 4o...
this heck can't even generate pictures, can't recognize another langueges, that's not a chatgpt
gonna reprot that space for disinformation for other people came ok
[UPD] Well, as expected that's just mistral - https://huggingface.co/spaces/KingNish/GPT-4o/blob/main/app.py
Apologies for the issue in image generation. The problem has been resolved, and image generation is now functioning properly.
The chat model is Idefics2 8b, a very powerful vision model with capabilities similar to top-notch vision models, and the voice chat model is Mixtral 8x7B instruct v0.1, which is more powerful than GPT 3.5 Turbo.
Please specify the language to bot if you wish to chat in another language.
Voice model currently only support english.
Apologies for the issue in image generation. The problem has been resolved, and image generation is now functioning properly.
The chat model is Idefics2 8b, a very powerful vision model with capabilities similar to top-notch vision models, and the voice chat model is Mixtral 8x7B instruct v0.1, which is more powerful than GPT 3.5 Turbo.
Please specify the language to bot if you wish to chat in another language.
Voice model currently only support english.
idefics is not powerful though, its a good model, but barely close to gpt4o. if even gpt3.5turbo.
Just rename this space to 'idefics-gpt4o-alternative' or something if you really want
idefics is not powerful though, its a good model, but barely close to gpt4o. if even gpt3.5turbo.
@Walmart-the-bag
Idefics 2 is also very powerful that it even surpassed GPT-4o and Llama3 70b in logical questions.
Question: On a fictional island, all the inhabitants are either knights, who always tell the truth, or squires (also called villains), who always lie. In the logic puzzle, a visitor comes to the island and meets some inhabitants. Usually, he has to deduce which type they belong to from the statements of the inhabitants, but sometimes he has to find out something else.
John: If and only if William is a squire, I am a squire.
Wilhelm: We are of different types. Use Boolean logic to solve the puzzle. Test all possible cases in a table. Remember: A squire must lie. So for his statement to be possible, it must be a lie. For a knight's statement to be possible, it must be true. Who is who?
(Correct answer is John is squire and William is knight)
(You can try asking with it any top-notch model most of them give wrong answer)
still not chatgpt4o.
stop givin' false information please and rename the space
The name "OpenGPT-4o" already stands out as different from "ChatGPT-4o."
@xxxmuzart This is just an attempt by one person to showcase that you can put together a system based on open stack. I don't find it's giving false information. It's a fun project. I appreciate the quick hack!
idefics is not powerful though, its a good model, but barely close to gpt4o. if even gpt3.5turbo.
@Walmart-the-bag
Idefics 2 is also very powerful that it even surpassed GPT-4o and Llama3 70b in logical questions.Question: On a fictional island, all the inhabitants are either knights, who always tell the truth, or squires (also called villains), who always lie. In the logic puzzle, a visitor comes to the island and meets some inhabitants. Usually, he has to deduce which type they belong to from the statements of the inhabitants, but sometimes he has to find out something else.
John: If and only if William is a squire, I am a squire.
Wilhelm: We are of different types. Use Boolean logic to solve the puzzle. Test all possible cases in a table. Remember: A squire must lie. So for his statement to be possible, it must be a lie. For a knight's statement to be possible, it must be true. Who is who?(Correct answer is John is squire and William is knight)
(You can try asking with it any top-notch model most of them give wrong answer)
I know and I agree with you, just saying this isn't gpt4o. It is a good project though.
Also: it's still not generating images some times
Thank you @1littlecoder and @Walmart-the-bag .
Also: it's still not generating images sometimes
@Walmart-the-bag the issue you reported should be fixed now, can you confirm it works as expected on your side?
Also: it's still not generating images sometimes
@Walmart-the-bag the issue you reported should be fixed now, can you confirm it works as expected on your side?
Works