CPU + 8 GB RAM
FP32 / quantized · FlagEmbedding
The roughly 2.27 GB checkpoint is CPU-capable; long sequences and batches add memory.
BAAI
Embeddings model
BGE-M3
A versatile multilingual embedding model for several retrieval styles.
Embeddings
Capabilities
What you can use it for.
The most useful reasons to choose this model, without making you read through its repository first.
Dense retrievalSparse retrievalMulti-vector retrievalMultilingual
01
Semantic search
02
Hybrid retrieval
03
Multilingual knowledge bases
For providers
Want to serve this model?
The provider requirements live here, separate from the information you need to choose and use the model.
Starting pointCPU · 8 GB RAM
Conservative estimate
Source and license
Check the original model.
OpenMayhem links back to the source repository so you can inspect the model card, files, license, limitations, and creator guidance directly.
Keep exploring