Rerank · conare-rerank-1

Put the right answer first.

conare-rerank-1 reads the question with each candidate passage and ranks them best first. It runs inside every Conare search, and on its own through a Cohere-compatible endpoint.

Book a demo

ViDoRe v3, reranking

70.2

nDCG@10, higher is better · #1 of 10

  1. 1conare-rerank-170.2
  2. 2Voyage rerank-2.568.5
  3. 3Voyage rerank-2.5-lite67.4
  4. 4zerank-267.0
  5. 5zerank-166.7

See the benchmark

Legal case search

1.3%

% ranked wrong, lower is better · #1 of 6

  1. 1conare-rerank-11.3%
  2. 2Claude Opus 5.52.5%
  3. 3GPT-6 Sol7.6%
  4. 4Mixedbread mxbai-rerank-v3.111.4%
  5. 5Contextual rerank-v253.2%

See the benchmark

Several conditions at once

2.5%

% ranked wrong, lower is better · #2 of 6, level with Claude Opus 5.5

  1. 1Claude Opus 5.51.1%
  2. 2conare-rerank-12.5%
  3. 3GPT-6 Sol3.8%
  4. 4Mixedbread mxbai-rerank-v3.116.7%
  5. 5Contextual rerank-v250.0%

See the benchmark

Each result has its own page with every model and how it was measured. All evals →

Searches with rules

Every condition, not most of them.

Real questions carry conditions: this court, that year, this clause. These benchmarks hide the one document that meets all of them among near-copies that each break one. conare-rerank-1 is level with Claude Opus 5.5 on both.

Legal case search

% ranked wrong, lower is better · 6 models

  1. 1conare-rerank-11.3%
  2. 2Claude Opus 5.5Level within the 95% interval2.5%
  3. 3GPT-6 SolLevel within the 95% interval7.6%
  4. 4Mixedbread mxbai-rerank-v3.111.4%
  5. 5Contextual rerank-v253.2%
  6. 6zerank-260.8%
How it was measured →

Several conditions at once

% ranked wrong, lower is better · 6 models

  1. 1Claude Opus 5.5Level with conare-rerank-1 within the 95% interval1.1%
  2. 2conare-rerank-12.5%
  3. 3GPT-6 SolLevel within the 95% interval3.8%
  4. 4Mixedbread mxbai-rerank-v3.116.7%
  5. 5Contextual rerank-v250.0%
  6. 6zerank-263.5%
How it was measured →

Using it

On by default. Or bring your own search.

Inside every search

Each query runs keyword and vector search, then conare-rerank-1 reorders the top passages before they come back. Metadata filters apply first, so it only sees documents you allow. It's on by default; pass rerank: false to skip it.

docs.query("how do refunds work?")

On its own

Already have a search stack? Send the query and up to 100 candidates to /v1/rerank, the same request Cohere's rerank API takes. They come back best first, each with its position in your list and a score.

POST https://api.conare.ai/v1/rerank

Documents per call
1 to 100
Document length
Up to 24,000 characters
Query length
Up to 8,192 characters
Endpoints
/rerank, and Cohere-compatible /v1/rerank and /v2/rerank
Rerank in the docs →
$ npm install conare
import { Conare } from "conare";

const docs = ["Orders ship in two days.", "Refunds last 30 days."];
const [best] = await new Conare().rerank("refund window", docs);
console.log(best);

Output

{ index: 1, score: 0.87, document: "Refunds last 30 days." }

See it on your own data.

Book a demo and we'll rerank your searches with you.

Book a demo