Hyper-v
Llama3 70B is HERE! Meta Outperforms Mistral8x-22b, Claude
Llama3 70B is HERE! 🦙🚀 Meta Outperforms Mistral8x-22b, Claude Sonet, 400B+ model soon?
#Llama3 #70B #Meta #Outperforms #Mistral8x22b #Claude
“Ai Flux”
Zuckerberg has delivered AGAIN with LLama3 8B and 70B. These models are incredibly powerful, LLama3 8B in certain cases actually outperforms LLama2 70B! What does this mean for the future of Meta and open source AI if models like this which are free and accesible to all can compete with GPT4…
source
To see the full content, share this page by clicking one of the buttons below |
Are you going to try to run it locally on the cluster and show us?!
Anxiously waiting for Llama 3 8X22B now!
it hallucinates more than mixtral 🙁
It might beat Gemini Pro 1.5 on some benchmarks. But after talking to both I have to say that Meta 3 general intelligence is significantly below GPT4, while Gemini 1.5 Pro is significantly above
"Even crazier is that they only trained with 16k H100s despite having ordered 600k. So only 2.6% of the compute from that order was used to create a GPT-4 level model. Wild times ahead." x: @Justin_Halford_
Zuck doesn't skip leg day.
I'm really liking command r+ because it just works really well and has a really long context length. I like it so much it has me seriously considering cancelling my chatgpt subscription.
I'm downloading llama3 rn and maybe the two together are enough to push me to cancel.
Hope we get a 14b soon
I think you kind of blew it by not clearing the context between questions. I'm not sure how much of an effect this has, but I imagine it's confusing to ask for flights and then right after, ask for bread recipes.
Also, it has a very limited 8k ctx so maybe that's why it choked on the tetris example.
The video and the questions were interesting and I appreciate your work a lot! Just this tiny issue might have compromised the performance a bit.
70B will run on 24gb vram ?
I think the context window is 9000. Not good
Surprisingly Meta has already shaken the industry. To be honest, if anyone had asked me 5 years ago which company would be the most open-source friendly in AI I would have added Meta at the bottom of the list.
I think I should switch from Sonnet to Llama 3 70B as a free chat assistant
good work like always
Those shots of zug "in the gym"… I feel like you should have done some kinda PSA first. "If there are children present…"
let's goooo 🎉
Incredible!! I'd even watch a video detailing this new 400B+ model we haven't seen yet!! 😮😮
Beats Mistral 7b… With 70b? Why the all so useless comparing to GPT4 ? That is more interesting. It is hard for me to understand why one would even want to put his own money into development something that is less then GPT4 to give it for free. Google at least having 1M parameters and interesting use cases.
Thank you Ai Flux! 🤗