Llama 405B-level performance, at a fraction of the cost
Meta’s Llama 3.3 multilingual LLM is an instruction tuned generative model in 70B that for some applications approaches the performance of Llama 3.1 405B and other frontier-level closed source models