Skip to content
ReviewRamble

The weekly brief

AI, reviewed, in your inbox.

  • New reviews of tools, models, MCP servers and repos, with the verdict up front
  • Re-checked prices and ratings when something changes
  • New analysis pieces as they are published

Models, agents, MCP servers and the tools actually worth your time.

One email a week. Unsubscribe with the link in any brief. How we use your email

Cross-encoder

A cross-encoder is a model that reads a query and a document together in one pass and outputs a relevance score. It is more accurate than comparing separate embeddings, but too slow to run over a whole collection.

Also called: bi-encoder, cross encoder

By Divya Sharma · Checked Nov 15, 2026 · How we review

The term matters most when choosing between two ways of using a transformer for search.

Bi-encoder vs cross-encoder

A bi-encoder encodes the query and each document separately into embeddings. Document vectors can be computed once and stored in a vector database, and a search is a quick vector comparison. This is how vector search works.

A cross-encoder feeds the query and a document into the model as one input, so every word of the query can attend to every word of the document. It produces a score, not a reusable vector. Nothing can be precomputed, so scoring a million documents means a million model calls.

The Sentence-BERT paper puts numbers on this: finding the most similar pair among 10,000 sentences took about 65 hours with a BERT cross-encoder, against about 5 seconds with its bi-encoder.

How they are combined

The usual pattern is retrieve and rerank: a bi-encoder or BM25 finds a few dozen candidates, then a cross-encoder rescores them. This is what most reranking models do, including mxbai-rerank v2.

Middle ground

Late interaction models such as ColBERT encode query and document separately but keep many vectors per text and compare them in a cheap final step, landing between the two approaches in cost and accuracy.

Related terms

Reranking
Reranking is a second search step that takes a short list of candidate documents and reorders them by relevance to the query, using a more accurate but slower model than the first search.
Late interaction (ColBERT)
Late interaction is a retrieval method, introduced by ColBERT, that encodes the query and each document separately into one vector per token, then scores a match by comparing those token vectors in a cheap final step.
Embedding
An embedding is a list of numbers (a vector) that represents a piece of text, an image or other data, produced so that items with similar meaning end up with vectors close together.
Vector search
Vector search finds items by comparing their embeddings to the embedding of a query and returning the closest ones, so results match on meaning rather than on shared keywords.
Hybrid search
Hybrid search runs a keyword search and a vector search on the same query and merges the two ranked lists, so results can match on exact words as well as on meaning.

Sources