As we continue to explore new post-training techniques, today we're releasing Llama 3.3 — a new open source model that delivers leading performance and quality across text-based use cases such as synthetic data generation at a fraction of the inference cost.

→ View original post on X — @aiatmeta,