Llama 3 400B: The AI Model That’s Redefining Open Source
Nothing is loaded from YouTube until you press play, and then without tracking cookies.
About this video
Llama 3 400B has just been released, and it's making waves in the AI community. In this short, we discuss the impact of Llama 3, its incredible training power, and how it compares to other open-source models. Hear expert insights on how this model stacks up against both closed models and graduate-level AI programs. If you're into AI and open-source innovation, don't forget to like, comment, and subscribe for more updates! #AI #Llama3 #MachineLearning #OpenSource #TechNews
Transcript
135 wordsCaptions as YouTube publishes them, usually machine-generated, so product and people names can be misheard.
- 0:00
I just saw that llama 3400b was released and so uh given you're a big proponent of Open Source and open weights I'm curious to hear kind of your reaction I think you know there have been these charts of Clos models going up like that and then open models not quite catching up but if you look at llama 405b the 405 billion parameter model so it's like 800 gabt of Weights fp16 16b Precision uh lower if you've got form of precision it was trained on 16,000
- 0:34
h100s for three months you know that's like probably double triple the computes of GPT 4 originally and So Meta has 350,000 h100s we got 35,000 now just for this that you've just had Elon announce his 100,000 h100s for kind of grop three and um