MLX FP16 format conversion of BAAI/bge-m3 at revision 5617a9f61b028005a4858fdac845db406aefb181. Converted on 2026-07-14T17:44:06.212531+00:00 with scripts/convert_bge_m3_to_mlx_fp16.py. The backbone, sparse_linear, and colbert_linear weights all come from that source revision. No additional training was performed.

Target runtime: MLX. Target dtype: FP16. The source model license is MIT; see the upstream model card for its complete terms and attribution.