I think this is a belated but smart move by OpenAI. They are basically fully moving in on Meta's strategy now, taking advantage of what may be a temporary situa

I think this is a belated but smart move by OpenAI. They are basically fully moving in on Meta's strategy now, taking advantage of what may be a temporary situation with Meta dropping back in model race. It will be interesting to see if these models now get taken up by the local model / fine tuning community the way llama was. It's a very appealing strategy to test / dev with a local model and then have the option to deploy to prod on a high powered version of the same thing. Always knowing if the provider goes full hostile, or you end up with data that can't move off prem, you have self hosting as an option with a decent performing model. Which is all to say, availability of these local models for me is a key incentive that I didn't have before to use OpenAI's hosted ones.