What models should be run?

[KennethEnevoldsen]

This poll is to get an idea about what models should be run. The hope is to create a minimal

Initial list:

• multilingual-e5-large-instruct
• multilingual-e5-{size}
• paraphrase-multilingual-mpnet-base-v2
• Commercial APIs (if sponsored)
  • text-embedding-3-{size}
  • embed-multilingual-v3.0

@MartinBernstorff, @imenelydiaker, @Muennighoff

[imenelydiaker]

I think text-embedding-3-large from OpenAI is expensive to run (time and money). The small would be fine.
But who will run the commercial API's models? If people already have API access it's okay, but I don't expect us to pass API keys to people, unless it's temporary..

Do we have a budget or something?