Isnt OLMO basically open like this? My understanding is people have recreated the model from the same data sources with repeatable responses, or reasonably close to the original (since LLMs never answer the same).