3 ms·
Ask HN: Why did Meta release OPT-175 and Llama separately?
In other words, why not keep making one model and improve it?
- breckenedge 4y agoMeta is huge. Probably released by different teams on parallel tracks. There can be benefits to diversity of effort.
- neximo64 4y agoThey're trained in completely different ways. Llama is trained in a way optimised for inference.