Reranker model
#7
by Duonglv - opened
Hello Microsoft team,
This is a great embedding model with only 0.6B params. It is stronger than many other models with similar size, especially qwen3 embedding 0.6B.
However, I wish that you guys will release a reranker 0.6B model as well. It’s very great if we use both embedding and reranker harrier-oss models togerther.
Beside that, running a big model needs much gpu vRam with very expensive cost, so it would be great if you guys can release variety model sizes, such as 0.3B, 0.6B, 1B, 2B, 4B,…
Thanks alot.