What are the best AWS services for faster text embeddings generation?

You might want to consider using Amazon Bedrock for your needs. It includes Titan Embeddings and models from Cohere and Anthropic, all accessible via API. It's designed to handle scale seamlessly, so you won’t have to worry about managing infrastructure. If you're looking for a SageMaker option, SageMaker JumpStart has 'text embedding' models with real-time endpoints that can provide lower latency compared to raw Hugging Face models. I’d recommend starting with Titan Embeddings on Bedrock since it's serverless and integrates well with other AWS services.

Have you checked out Amazon Bedrock? It provides embeddings from both AWS and Cohere. Some users say Cohere has better performance, but it might come at a higher cost. Here's a couple of links to get you started: [Bedrock User Guide](https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-embed-v4.html) and [Titan Models](https://docs.aws.amazon.com/bedrock/latest/userguide/titan-embedding-models.html). You could also explore using Bedrock Knowledge Bases for creating cheaper retrieval-augmented generation (RAG) setups with embeddings.