BGE-small
BGE-small is a lightweight text embedding model open-sourced by the Beijing Academy of Artificial Intelligence (BAAI). It is the smallest-parameter variant in the BGE model series, with a model size of approximately 24 MB. Its fast inference speed makes it well suited for local or edge deployment.
Timeline (last 90 days)
In experiments based on the Banking77 dataset, the BGE-small + logistic regression combination achieved 93.3% accuracy, while Jev achieved only 83.2%.
The BGE-small + logistic regression solution requires only approximately 9ms to run locally.
BGE-small is a lightweight text embedding model open-sourced by the Beijing Academy of Artificial Intelligence (BAAI), is the smallest-parameter version in the BGE series, and has a model size of approximately 24MB.
All Facts (3)
In experiments based on the Banking77 dataset, the BGE-small + logistic regression combination achieved 93.3% accuracy, while Jev achieved only 83.2%.
50%UnverifiedThe BGE-small + logistic regression solution requires only approximately 9ms to run locally.
50%UnverifiedBGE-small is a lightweight text embedding model open-sourced by the Beijing Academy of Artificial Intelligence (BAAI), is the smallest-parameter version in the BGE series, and has a model size of approximately 24MB.
50%