Static Retrieval MRL EN V1
sentence-transformers/static-retrieval-mrl-en-v1
published Oct 2024 · updated Jan 2025
Static Retrieval MRL EN V1 is a static embedding model that maps sentences and paragraphs to a 1024-dimensional dense vector space using pre-computed token embeddings, optimized for semantic search with Matryoshka representation learning.
specs
| Task | Embedding / Semantic Search |
| Architecture | StaticEmbedding (EmbeddingBag with BERT uncased tokenizer) |
| Parameters | 0 active parameters (pre-computed token embeddings) |
| License | Apache 2.0 |
| Output Dimensionality | 1024 (truncatable via Matryoshka) |
| Similarity Function | Cosine Similarity |
best for
- ·Semantic search and document retrieval at scale
- ·Fast embedding generation on CPU with minimal resource usage
- ·Use cases requiring flexible embedding dimensionality via Matryoshka truncation
FAQ
Semantic search and information retrieval where speed and low resource usage are critical. It is designed for efficient nearest neighbor search.
It is 100-400x faster on CPU and 10-25x faster on GPU while achieving 87.4% of its retrieval performance.
Apache 2.0.
Yes, the model was trained with Matryoshka loss, allowing you to truncate the embedding dimension (e.g., to 256) with minimal performance loss. Use the truncate_dim parameter.
Use the OpenAI-compatible endpoint with your API key. Refer to gigarouter documentation for endpoint details.
We're benchmarking and onboarding Static Retrieval MRL EN V1 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.