Mxbai Rerank XSmall V1
mixedbread-ai/mxbai-rerank-xsmall-v1
published Feb 2024 · updated Apr 2025
Mxbai Rerank XSmall V1 is a cross-encoder rerank model that reorders documents based on relevance to a query, optimized for efficiency and high performance at a small size.
specs
| Task | Reranking |
| Architecture | CrossEncoder (DeBERTa-v2 backbone) |
| License | Apache 2.0 |
| Formats | ONNX, Safetensors |
about this model
mxbai-rerank-xsmall-v1 is a CrossEncoder reranker model based on the deberta-v2 architecture, designed to reorder search result lists according to predicted relevance to a given query. It is trained on real-life search queries paired with top-10 search engine results, using a large language model to generate relevance labels.
Key strengths
The model is capacity-efficient and delivers high performance at a very small size, making it suitable for production reranking pipelines where latency and resource constraints matter. It is part of a three-model family that also includes mxbai-rerank-base-v1 and mxbai-rerank-large-v1, allowing users to trade off size and performance as needed. The model is available under the Apache 2.0 license and is supported in ONNX and Safetensors formats, as well as via Transformers.js for JavaScript or Node.js environments.
best for
- ·Reordering top-N search results for retrieval-augmented generation (RAG)
- ·Improving precision of document retrieval in enterprise search
- ·Relevance scoring for question-answering systems
FAQ
It excels at reordering a small set of candidate documents by relevance to a query, making it ideal for RAG pipelines, enterprise search, and QA systems.
It is capacity-efficient and high-performing at a very small size, offering good performance with a slight increase in non-relevant result scores. It is the smallest model in a three-model family (xsmall, base, large).
It is released under the Apache 2.0 license, allowing free use, modification, and distribution.
It accepts a query and a list of documents (or query-document pairs) and outputs relevance scores for each pair. The hosted API uses a JSON payload with a query and documents array.
Send a POST request to the gigarouter OpenAI-compatible endpoint with your API key, specifying the model name and the query/documents in the request body.
We're benchmarking and onboarding Mxbai Rerank XSmall 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.