Snowflake Launches Text-Embedding Model for Retrieval Use Cases

[rawgabbit]

Quickstart: https://quickstarts.snowflake.com/guide/asking_questions_to_...

[unraveller]

Alibaba also just released gte-large-en-v1.5 a 434M 1.62GB RAM model scoring 57.91 on retrieval. Beating out the snowflake-large by 2 whole points for +400MB extra RAM.

https://huggingface.co/spaces/mteb/leaderboard

looks like we are in for a plataeu with the small embed models with no real performance jumps over the last months. snowflake-arctic-embed-medium being the sweetspot at 410M RAM scoring 54.91

[dar8919]

Wow, shows very good performance on my wikipedia dataset. Incredible that companies are open sourcing so much good stuff. Hope this trend continues.