Zuck: To give a sense of performance, this 8B model is nearly as good as the biggest Llama 2 model. This 70B model is around 82 MMLU with leading reasoning and math benchmarks. The 400B+ model is currently around 85 MMLU but it's still training, so we expect it to lead on several benchmarks when it's ready. More details on the two 24K H100 GPU clusters they used to train Llama 3.

I don't typically upvote for major corporations on Product Hunt. I like to use this site to let smaller teams get more visibility but I had to give an upvote to Llama 3. Llama is what I wish all the other LLMs would be: open source. Kudos to Meta for leading the charge on this.