We’ve fixed all known issues with built-in embedding models… The remote embedding models (openAI specifically) has issues we’re working on