Yet Another Farcical Self-Delusion? I do not work with text embedding models currently, but such threads are utterly hilarious: https://twitter.com/Nils_Reimers/status/1487014195568775173?s=20&t=z8jAsiDgoASIOqppzhNnkQ If anyone does, please explain if this is biased. But when I worked with such models ... low key public models published by FAIR / Google worked decently, so idk. If OpenAI in reality is as useful as Tesla car service ... well you know =) I can only add that when we were looking for some base compact multi-language transformer model for fine-tuning ... the best we found was dated ~2019, which I find fucking hilarious. But ofc there were several people re-uploading the most popular models from 2018 by the hundreds... claiming to make them more compact ... just by cutting unnecessary embeddings for a given language. #no_bs