Skip to content

Category

Search

Retrieval-augmented and web-grounded models. Models and rankers that ground answers in retrieved evidence. Reach for these when an answer has to be traceable to a source. A pass over results usually lifts a retrieval stack more than a bigger generator does, and it is priced per thousand queries rather than per token.

1 model in Search from 1 provider

R/1K queries Rerankers are billed by the thousand queries, so the cards below show a query price rather than a token price.

Top model

Where this category stands today

View all →
Rerank 3.5 by Cohere
CohereOn request

Rerank 3.5

cohere/rerank-3.5

The retrieval quality multiplier, reorders candidate documents by true relevance, per query.

SearchBatch

4K ctx · Price on request

Browse all search models →

1 model in the directory

What to look for

What to look for in Search

One model tracked today. These are the criteria that will separate the next ones.

  1. 01

    Rerankers multiply recall

    A reranking stage often improves RAG quality more than upgrading the generator.

  2. 02

    Query pricing

    Rerank models price per 1K queries, a different economics from token pricing.

  3. 03

    Freshness requirements

    Web-grounded answers need live retrieval; static corpora don't.