BGE Base EN v1.5
Xenova/bge-base-en-v1.5
published Sep 2023 · updated Jul 2025
BGE Base EN v1.5 is an embed model that converts English text into 768-dimensional dense vectors for semantic search, clustering, and classification.
specs
| Task | Feature Extraction / Text Embedding |
| Architecture | BERT-base (BGE) |
| Output Dimensions | 768 |
| License | MIT |
about this model
Xenova/bge-base-en-v1.5 is an English text embedding model optimized for retrieval and semantic similarity tasks, converted to ONNX weights for efficient web deployment and hosted as a managed API by gigarouter.
Part of BAAI’s General Embedding (BGE) series, the v1.5 update alleviates similarity distribution issues and enhances retrieval capability without requiring an instruction prefix (a recommended query prefix can still be used for retrieval tasks). The model outputs 768-dimensional embeddings and is released under the MIT license.
Benchmark Performance
Evaluated on the MTEB benchmark, the model achieves competitive scores across classification, retrieval, and semantic textual similarity tasks:
| Task | Metric | Score |
|---|---|---|
| ArguAna Retrieval | NDCG@10 | 63.61 |
| AmazonPolarityClassification | Accuracy | 93.39 |
| BIOSSES STS | Cosine Spearman | 86.94 |
| Banking77Classification | Accuracy | 86.95 |
These scores reflect the model’s strong general-purpose embedding quality for English text, particularly for dense retrieval and classification workloads.
best for
- ·Semantic search and passage retrieval from a document corpus
- ·Sentence similarity and clustering for English text
- ·Text classification (e.g., Amazon review polarity)
FAQ
It accepts English text strings and outputs 768-dimensional dense vectors when using mean pooling and normalization.
It uses a recommended query prefix: "Represent this sentence for searching relevant passages: " for retrieval tasks.
MIT.
Use the gigarouter OpenAI-compatible endpoint with your API key; set the model parameter to "Xenova/bge-base-en-v1.5" and send your text in the input.
768.
We're benchmarking and onboarding BGE Base EN v1.5 as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.