Skip to main content

BGE-M3

Compare

bge-m3

Multilingual retrieval model supporting dense embeddings, sparse retrieval and multi-vector retrieval across more than 100 languages, with up to 8192 input tokens. Official weights and FlagEmbedding inference code are available under MIT.

Context window
8.2K
Max output
-
Input
Output
Released
—
Weights
Open
License
MIT
Capabilities
Open weight

Official list prices

No official pricing available.

Performance

No evaluation results available.

Providers

OutputCache Readvs Official
DigitalOcean---—
NovitaAI---—
OpenRouter--—

Page 1 of 1

Alternatives

—33K---
2025-06-03
—8.2K---
-
-
—8.2K$0.1/M--
-
—32K---
-
—41K---
-
—32K--
-
—8.2K$0.02/M--
-
—512---
-

ModelRegistry · Catalog updated 2026-11-16

Popular Model Organizations

The organizations behind the models, ordered by model count.

All 95 organizations