HACKER Q&A
📣 behnamoh

Why did Meta release OPT-175 and Llama separately?


In other words, why not keep making one model and improve it?


  👤 breckenedge Accepted Answer ✓
Meta is huge. Probably released by different teams on parallel tracks. There can be benefits to diversity of effort.

👤 neximo64
They're trained in completely different ways. Llama is trained in a way optimised for inference.