We are finding there is a trade-off between model performance and hosting costs post-training. The optimal outcome is where we have a model that performs well o

We are finding there is a trade-off between model performance and hosting costs post-training. The optimal outcome is where we have a model that performs well on next-token prediction (and some other in-house tasks we've defined) that ultimately results in a model that we can host on the lowest-cost hosting provider rather than be locked in. I think we'd only go the proprietary model route if the model really was that much better. We're just trying to save our selves weeks/months of benchmarking time/costs if there was already an established option in this space.