OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
RAG & Knowledge Systems

Flashrank

Flashrank is an open-source Python library that reranks search and retrieval candidates using pairwise, pointwise or listwise models. Developers can place it between retrieval and downstream LLM processing to reorder passages by relevance.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Flashrank — OpenViST project profile in RAG & Knowledge Systems.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Flashrank when a Python search or retrieval pipeline already produces candidates and needs a reranking step before results are consumed by an LLM or another downstream stage.

Who it is for

  • Python developers maintaining search or retrieval pipelines.
  • Teams building RAG workflows that need a reranking stage before LLM input.

Documented uses

  • Reorder retrieved passages for a query before passing them to an LLM.
  • Add pairwise or pointwise cross-encoder reranking to a search pipeline.
  • Choose a listwise LLM-based reranker for candidate ordering.
Before you choose

Requirements & limits

The repository identifies Flashrank as open source under the Apache-2.0 license. The README documents installation from Python package commands.

  • The default pairwise package is installed with pip install flashrank; LLM-based listwise support uses the flashrank[listwise] extra.
  • The README lists models from about 4 MB to about 4 GB, so model choice affects footprint.
  • The README says query and passage token lengths and model depth influence reranking time; an oversized max_length can slow processing.
  • The listed MultiBERT model supports more than 100 languages, and its README description advises against using it for English.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install Flashrank for pairwise reranking

Add the default lightweight pairwise reranker package to a Python project.

Before you start

  • A Python environment with package installation available.

Steps

  1. Install the default package with the command below.

    Documented command
    pip install flashrank

  2. Import Ranker and RerankRequest from flashrank.

  3. Create a ranker, for example with Ranker(max_length=128); adjust the maximum length to fit the query and passage lengths.

Expected outcome

The Python project has the documented default pairwise reranker package and a configured Ranker instance to use in a retrieval pipeline.

Guide limitations

  • The README notes that an unnecessarily large max_length can increase response time.
Documentation 1 ↗
Source-based practical guide

Install Flashrank with listwise reranker support

Install the package extra for LLM-based listwise rerankers.

Before you start

  • A Python environment with package installation available.

Steps

  1. Install the listwise extra with the command below.

    Documented command
    pip install flashrank[listwise]

  2. Consult the repository's supported-model information to select a listwise model.

Expected outcome

The Flashrank package is installed with the documented listwise support extra.

Guide limitations

  • Listwise rerankers are described as LLM-based and may have different model-size and token-length requirements.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
9 October 2026
Original website title
PrithivirajDamodaran/FlashRank · Official homepage title · retained in its source language
How we review sourcesSuggest a correction