Metas LLAMA 405B Just STUNNED OpenAI! (Open Source GPT-4o)

00:00 - Llama 3.1 announcement
03:25 - 405B model benchmarks
05:41 - 8B and 70B model updates
06:49 - Human evaluations
07:48 - Architecture choices
08:51 - Multimodal capabilities
10:01 - Vision performance
11:00 - Video understanding
11:50 - Audio features
12:53 - Tool use demo
13:45 - Future improvements
14:04 - Accessing Llama 3

Links From Todays Video:

Welcome to my channel where i bring you the latest breakthroughs in AI. From deep learning to robotics, i cover it all. My videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on my latest videos.

Was there anything i missed?

#LLM #Largelanguagemodel #chatgpt
#AI
#ArtificialIntelligence
#MachineLearning
#DeepLearning
#NeuralNetworks
#Robotics
#DataScience

[OceanGateEngineerHire]

Zuckerberg's redemption arc has been magnificent to see. Good for him.

[viddarkking]

128K Context Windows Hum.
Well looks like the RP Sessions are about to get a lot more fun.
Not to mention the shear length that the character cards will now be able to have.
I may just jump back into Sillytavern for a bit once the Fine tones are out.

[dr.emmettbrown]

It is NOT Open-Source but Open-Weights.The difference is essential.

[jonasmuller]

Nice way of kicking Out the competition

[TManjen]

Hi. You can also use it in the UK via Perplexity Ai. As a pro user you can select underlying model and 405b is there already. Thanks.

[massiah]

the real time image creation is so cool... just type imagine and it works seamlessly

[jeremylane]

one thing failed to mention on that graph 6:30, is that 3.1 8b matches or beats mixtral/gpt4.5t

[zyzhang]

Looks like diminishing returns hit pretty hard (3.1 70b is not that far off)

[charleshopper]

This is the smartest thing that Meta has done hands down. Looking forward to using the new LLM's.

[elawchess]

It's on huggingface chat as well but it gets overloaded from time to time

[thorvaldspear]

playing this in the background to help me wait until AI Explained posts

[pressrepeat]

I tested this 405B model against the 3.0 model (both 70B and 8B) and for some reason the new model only uses data until 2020, while the 3.0 model used data until the end of 2022. Weird.

[HitsInSandbox]

I tried the 8B and wow what a difference in speed and accuracy. After the test I have removed 80% of my other LLM I had on the system why waste space if I can do all them in just one LLM with overall better on all levels.

[redthunder]

1:38 this is HUGE. essential function calls native to the model.

[GonzoWasHere]

Why do the “people” in this vid seem like terminators on “friendly mode”?

[MrDudukmakarna]

I copied the link of video and pasted it to chat gpt to summarize it. Thanks.

[viddarkking]

A question I am really looking forward to have answers is if we can use Llama 3.1 70B on a RTX 5090.
Because that would be a true paradigm shift.

[ThomasConover]

What hardware do I need to buy to run the 405B model

[gabrielefilosofi]

all these bunch of nuclei and interconnecting pathways, do they preserve the somatotopic specificity of the primary motor cortex?

[zorororonoa]

After watching the video I am pretty pretty stunnned.