Our new 8B and 70B parameter Llama 3 models establish a new state-of-the-art, with improvements in pretraining and post-training, reducing false refusal rates, enhancing alignment, diversity, reasoning, and more.
Preference rankings show strong performance of our 70B instruction-following model in real-world scenarios, establishing a new state-of-the-art for LLM models.
Collection
[
|
...
]