Why has the 2026 default become wide retrieval plus a cross-encoder rerank?

Attempt it

The two stages optimize different things. Stage 1 (dense or hybrid retrieval) prioritizes recall: a fast bi-encoder index pulls a generous candidate set so the right answer is in the pool, even if it is not at position 1. Stage 2 (cross-encoder rerank) prioritizes precision: a smaller, more expensive model scores each candidate against the query jointly, sharpening the ranking so the top-3 are actually the best matches. Doing precision in one stage is too slow; doing recall in one stage is too imprecise. The handoff lets you tune k1 (recall) and k2 (precision) independently. Cohere Rerank 3.5, BGE Reranker v2, and Voyage Rerank-2 are the common 2026 rerankers.

Your answer

Quick reference

TL;DR

Stage 1 is a fast bi-encoder optimized for recall over a wide candidate pool; stage 2 is a slow cross-encoder that scores query and chunk jointly to sharpen precision on the survivors.

Memory aid

Sign in for the mnemonic that makes this stick. See Pro pricing.

Easy to grasp

Picture hiring for one open seat at a software company. You do not interview every applicant in depth. First, a resume scan sweeps thousands of resumes and pulls maybe fifty that look plausible. That sweep is fast and forgiving on purpose: you want the right candidate in the pile, even if they are not ranked first. Then a smaller panel does deep interviews with those fifty and picks the actual top three. Two passes with two different filters do a better job than one pass with one filter that has to be both fast and thorough. RAG retrieval works the same way. Vector search is the resume scan. The reranker is the interview panel.

Key concepts

Context engineering Retrieval-augmented generation Embeddings Vector databases

Quick scan · 2–4 min

Everything important, quickly.

4 minutes: bi-encoder versus cross-encoder, recall versus precision split, k1 and k2 defaults, three 2026 rerankers, one sizing question for sanity check.

Code example

python

from anthropic import Anthropic
import cohere

vector_db = ...  # Faiss / Pinecone / Qdrant client
co = cohere.Client()

def retrieve(query: str, k1: int = 30, k2: int = 5) -> list[str]:
    # Stage 1: wide bi-encoder recall
    candidates = vector_db.search(query, top_k=k1)
    # Stage 2: cross-encoder rerank to top k2
    docs = [c.text for c in candidates]
    rerank = co.rerank(model="rerank-3.5", query=query, documents=docs, top_n=k2)
    return [docs[r.index] for r in rerank.results]

Real-world examples

Real products, models, and research that use this idea.

Cohere Rerank 3.5 is the default reranker for many production RAG stacks in 2026; pairs naturally with Cohere Embed v4 or any dense index.
BGE Reranker v2 (from BAAI) is the open-weight 2026 baseline; runs well alongside BGE-M3 embeddings.
Voyage Rerank-2 from Voyage AI focuses on technical and code domains.
Anthropic's published RAG cookbook for Claude Opus 4.7 recommends top-50 dense retrieval plus a cross-encoder rerank to top-5 before context assembly.
Pinecone, Weaviate, and Qdrant all ship native rerank integrations in 2026, treating the two-stage pipeline as the assumed shape.

Interviewer follow-ups

What an interviewer would ask next. Try answering before peeking at the approach.

QHow would you tune k1 and k2 for a high-recall medical-literature search versus a low-latency support bot?

Sign in or go Pro to see the approach for this follow-up.

QWhen does hybrid (BM25 + dense) retrieval in stage 1 beat pure dense?

Sign in or go Pro to see the approach for this follow-up.

QHow do you evaluate the rerank stage in isolation?

Sign in or go Pro to see the approach for this follow-up.

60-second cheat sheet

The night-before-the-interview bullets. Scan these on the way to the call.

Sign in for 5 unlocks/day or go Pro for unlimited revision sheets. See Pro pricing.

References

Primary sources. Skim if you want the original framing.