GPT-4 Surpassed Claude 3 (Again) | GPT-4 Turbo fully tested

Sdílet
Vložit
  • čas přidán 11. 07. 2024
  • FULLY Tested GPT-4 Turbo, the flagship model from OPENAI.
    Benchmark of GPT-4 vs GPT-4 Turbo
    _______ 👇 Links 👇 _______
    lmstudio.ai/
    🤝 Discord: / discord
    💼 𝗟𝗶𝗻𝗸𝗲𝗱𝗜𝗻: / reda-marzouk-rpa
    📸 𝗜𝗻𝘀𝘁𝗮𝗴𝗿𝗮𝗺: / redamarzouk.rpa
    🤖 𝗬𝗼𝘂𝗧𝘂𝗯𝗲: / @redamarzouk
    www.automation-campus.com/
    _______ 👇 Content👇 _______
    00:00 Introduction and AI Week News
    00:11 Launch of a new AI model by Mistral
    00:30 New Google AI products announcement
    00:55 GPT-4 Turbo update via ChatGPT Plus
    01:24 Tiny_Benchmark Presentation
    02:00 Math and Reasoning 1
    02:11 Math and Reasoning 2
    03:13 Math and Reasoning 3
    03:35 Math and Reasoning 4
    04:34 Logic and Reasoning 1
    05:00 Logic and Reasoning 2
    05:44 Logic and Reasoning 3
    06:30 Logic and Reasoning 4
    07:13 Logic and Reasoning 5
    07:37 Coding 1
    08:02 Coding 2 (minesweeper game)
    12:21 Dilemma
    13:52 Final Results (Tiny_Benchmark)
  • Věda a technologie

Komentáře • 15

  • @beyondthebounce23
    @beyondthebounce23 Před 2 měsíci +4

    I pay for both of these. I use them for coding. I can tell you right now that GPT4 Turbo does not even come close to being as good as Claude 3 Opus. Heck I find even Claude 3 Sonnet to be better.

  • @gabrielmestre3623
    @gabrielmestre3623 Před 2 měsíci +1

    I really need help because I have the plus and even so in the playground I only have the 3.5 models.. besides that I don't notice any difference within the gpt chat

    • @ika9
      @ika9 Před 2 měsíci +1

      U have to understand gpt4 web is different than gpt4 playground api even u have plus sub u need to subscribe to api aswell

    • @redamarzouk
      @redamarzouk  Před 2 měsíci +1

      Hello Gabriel,
      First for the the chatgpt website (chat.openai.com) if you select GPT-4 then you're using the model (gpt-4-turbo-2024-04-09) which is the latest and best model out there compared to even claude opus, so maybe you didn't notice the difference but it's better.
      Second for accessing gpt-4 models in playground, you need an api subscription (at least a 5$ one time payment for api calls) and that will give you the additional models.
      hope that answers your question.

  • @giliyehuda
    @giliyehuda Před 2 měsíci +1

    Is the new GPT update just found in PLAYGROUND

    • @redamarzouk
      @redamarzouk  Před 2 měsíci

      It's available in the Chatgpt website as well starting from last week (considering you have GPTPLUS subscription).

    • @giliyehuda
      @giliyehuda Před 2 měsíci

      @@redamarzouk Thank you very much for the answer, how do I know if my GPT is updated or not, what is the index?

    • @redamarzouk
      @redamarzouk  Před 2 měsíci +1

      @@giliyehuda if you're using on chat.openai.com/ website then it's update by default, so no action is needed on your behalf.
      if you're calling it in a python script you will have to define the exact version of gpt-4 you want to call, in this case the latest and best model is
      gpt-4-turbo-2024-04-09
      it's the model I've reviewed in the video, it's has 128K context window and updated up to April 9th of this year.

  • @bobharris5093
    @bobharris5093 Před 2 měsíci +1

    1 day and these news are already obsolete 😂

    • @redamarzouk
      @redamarzouk  Před 2 měsíci +1

      I gotta agree 1 day is like 1 month long in AI

  • @dg-ov4cf
    @dg-ov4cf Před 2 měsíci

    whaat if the helicopter is maade off OATMEAL :D

    • @redamarzouk
      @redamarzouk  Před 2 měsíci

      I once asked I dunno if it was gemma or another model on how many helicopters we can eat and the answer was 1.

  • @devilsolution9781
    @devilsolution9781 Před 2 měsíci

    So theyre downgrading to upgrade, seems like apples old tricks

    • @redamarzouk
      @redamarzouk  Před 2 měsíci

      That's what I felt using it but maybe that gpt-4 version is the very first one that wasn't good anyways (that's the problem with closed source you don't get the whole information) .

    • @devilsolution9781
      @devilsolution9781 Před 2 měsíci +1

      @@redamarzouk nah it makes sense from a financial perspective, they arent improving at a fast enough rate so they dowgrade the lower model and put those features on the top model. I dont know how much electricity the compute costs but no doubt thats how they condone their policy.